Internal and External Collaborative Shaping: The Role of Official Information and Online Communities in Shaping a City’s Image

Abstract

City image is essential for city marketing, yet the impact of “outside–in” shaping by social media in the Web 2.0 era has been largely overlooked. The decentralized and diverse Web 2.0 environment now dominates online information dissemination, influencing not just cyberspace, but also the physical urban landscape. These externally driven city images increasingly reflect and interact with traditional “inside–out” images shaped by official sources. Understanding the influence of external actors via social media compared to traditional internal sources, like government websites, is crucial. This dual analysis offers insights into city image formation, helping cities refine their marketing strategies. This study analyzed a representative social media platform alongside official government websites, using programming and a naive Bayes classifier. We developed a method to categorize the city images of selected U.S. world cities based on these two different media sources. The results are as follows: (1) We establish a city image categorization system that divides the considered U.S. world cities into four and five categories based on social media and official government website content, respectively. (2) We compare the groups and logics shaping global city images in different cyber eras based on the example of the U.S. world cities, and based on this, we explore the relative roles of groups outside the city. (3) We identify the preferences of forming different city images between external groups based on social media and internal forces based on government websites. In summary, this article takes world cities as an example to demonstrate that, in the Web 2.0 era, the image of a city depends on both internal and external groups and has varying degrees of preference. The unique urban image of each city is formed through two media content streams and quantitative preference.

1. Introduction

During the 1980s, human society entered the era of globalization, and all cities needed to compete with other cities around the world for advantageous resources, such as capital, talents, and technology, and for improving their rankings in the world’s urban system [1] (pp. 407–413). In the 1990s, academics defined city marketing as the act of “packaging, upgrading, and creating a variety of resources that have the potential to attract investment, elites, and tourists, and selling them through the medium of marketization in order to enhance the attractiveness of the city to advantageous resources” [2,3]. City image is the perception of a city by an individual or a group that can be expressed with pictorial and textual means and is important for city marketing and enhancing a city’s international competitiveness. Kevin Lynch’s book The Image of the City, published in 1960, is regarded as a classic in the theories of city image, which triggered a surge of research on city image. Since then, many scholars have perfected and deepened this field on the basis of Kevin Lynch’s research. In recent years, several scholars have attempted to propose conceptual systems for urban spatial images, and initially explored the spatial characteristics, elements, spatial patterns, and sense of place of images, contributing to shaping urban characteristics and spatial quality through urban planning and management, urban landscape design, and urban tourism development [4].

Digital and, in particular, internet technologies are pervasive in our daily lives, shaping our interactions and experiences. With their rapid advancement, the twenty-first century has been heralded as an era of profound transformation and reform. This is evident in the way in which communication and information are accessed, revolutionizing how we connect and engage with the world around us, which directly changes the process of forming a city’s image in people’s minds [5] (pp. 181–198). Today, the internet allows us to connect anytime and anywhere, granting us instant access to a wealth of information. Initially introduced through websites, this technology was dubbed Web 1.0, which emphasizes its role as a one-way conduit for disseminating information [6,7,8,9] (6, pp. 181–198; 7, pp. 211–225; 8, pp. 233–259). In this era, urban marketing research typically revolved around urban governments. At the beginning of the 21st century, with the entrepreneurial transformation of governments, the role of city governments in the development of urban globalization was increasingly emphasized, and urban marketing scholars thus began to pay attention to the study of city governments as the main body of marketing [10,11] (10, pp. 489–521; 11, pp. 1177–1186). Recognizing the limitations of this one-way information flow on websites, efforts to innovate led to significant changes in Web technologies. This ushered us into the era of Web 2.0, characterized by the emergence of tools, resources, and environments that support greater user engagement. This marked a pivotal shift toward technologies that facilitate more interactive and participatory experiences on the internet [12] (pp. 93–106). As a result of this paradigm shift, website owners and users have become active participants in Web content design and creation [13], where the flow of new information is from the “outside–in”. Moreover, internet information and data can comprehensively reflect the nature of cities [14] (p. 667), meaning that urban functions can be recognized from this perspective.

In the new era of decentralized networks, much of the discussion on city image focuses on changes in the power of discourse/information, which in turn lead to changes in the structure of network content production. Network content is no longer one way, from top to bottom, but is strongly decentralized; in the era of Web 2.0, this decentralization is mainly manifested in the production and processing of online content by the public on social media. However, few studies have focused on the differences in the demographic scope of the groups communicating through official and social media channels [15,16] (15, pp. 1–7; 16, pp. 353–356). The significance of this study lies in the exploration of the interplay between city image formation from the outside–in, based on social media big data, and image construction from the inside–out, based on official government websites. This dual approach allowed us to address a key gap in the existing urban research, which often lacks integration and fails to establish strong and comprehensive correlations with media data. By comparing these two modes of image shaping, this study aimed to provide a more accurate and comprehensive understanding of city perception. This bi-directional perspective analysis is essential for creating a nuanced assessment of city image, contributing to a more informed urban planning and decision-making process that reflects both the city’s internal communication and the external environment. In addition, with regard to the duality of “inside–out and outside–in” image shaping proposed here, the existence of preferences between the two is also important for the identification of a particular city image.

In this study, the online community content production process was regarded as a reproduction of the city entity in the external environment. We chose X, one of the most representative social media platforms in the United States, as a source of data and combined methods of content mining, data collection, and cluster analysis. The aim was to identify different cities based on the contents of official government websites and social media and analyze the correspondence, differences, and preferences of the city characteristics made available in the two media environments, so as to supplement the human geography study of city image and provide a reference for cities to enhance their online brand image. We confirmed that online communities dominate the formation of city image in the Web 2.0 era, together with the Web 1.0 inside–out information system, which makes up for the shortcomings of the theory of city image formation based on public information in the new era. Quantitative methods are sought to identify the unique image of each city from two types of media content, with the aim of avoiding the homogenization of the city’s image and establishing a more credible and accurate brand image for each city in a context where marketing goals are essentially the same. We attempted to address the following questions by using an online platform: What methods can be used to identify a city’s image through social media and how does this image differ from and relate to traditional official city marketing? Is there a difference in scope between the groups communicating through social media and official channels, and what is the significance of this difference for the city image? Is there a preference for social media and official channels for the finalization of the city’ image, and can its degree be identified?

The remaining parts of this article include Literature Review Section, which presents relevant previous research, methods, and a theoretical basis, and explains why American world cities were chosen as samples; Data and Methods Section, detailing the methodology and data processing details used in this article; Results Section, demonstrating the specific results and analysis of identifying the city image and dual data source preference in a quantitative manner; Conclusions Section, summarizing the specific contributions and theoretical innovations of this article; and Discussion Section, where we discuss the possibility and feasibility of future research, while calling for attention to the limitations of this article and advocating for research to supplement new theories.

2. Literature Review

In 2004, the concept of Web 2.0 emerged during a collaborative brainstorming session involving MediaLive International and Tim O’Reilly, an American publisher renowned for his focus on emerging technologies and networks [6,17,18,19,20] (6, pp. 181–198; 18, pp. 1217–1221; 19, pp. 17–37). Web 2.0 is distinguished from Web 1.0 by its enhanced writing capability. As a result, some researchers define Web 2.0 as “read/write”, highlighting its interactive nature compared to the primarily static content of Web 1.0 [6,7]. The definition of Web 2.0 is highly debated, with various interpretations. However, it generally refers to an environment where individuals can create and modify content, reorganize information according to their needs, and participate in collective intelligence. User-generated content and social media platforms, including blogs and social networking sites, have become emblematic of the Web 2.0 environment [21] (pp. 214–226). Research on Web 2.0 has gradually formed a recognized concept of online community. The term “online community” refers to the social aggregation generated from the social internet (social media), that is, the personal relationship network formed in virtual space by people engaging in public discussions for a long time based on sufficient human emotions [22]. Since the emergence of online communities, the concept has been of interest to the human geography community as a relational bridge between virtual and real space (especially urban space) [23,24] (23, pp. 601–606; 24, pp. 2075–2086). Online communities are becoming a platform for the production, distribution, and discussion of information related to urban development, as well as for cross-city exchanges among users, thereby expanding the Chinese public’s understanding of the city to include an understanding of and participation in China’s new urbanization process from an interurban perspective that transcends the individual scale. Therefore, human geographers should pay attention not only to “objectively existing” geographic space, but also to “truly existing” virtual space [25] (pp. 1894–1901).

Research in the Web 1.0 era on city image as reflected in media messages has not yet covered much social media, focusing mainly on image marketing conducted by official government media. To date, city images, as expressed on official government websites, have been studied in two ways. The first category includes content-oriented research on image building. The analysis of images from the government websites of 12 British cities reveals a focus on publicizing good social integration as part of the atmosphere of these cities through images showing non-White populations, women’s employment, and daily life [26] (pp. 13–31). Second, the relevance factor of diversified content was studied. The city government can more easily use the internet, as a virtual medium, to filter content to present a chosen city image to the outside world in line with its will. Some scholars believe that city governments may choose web content based on the economic, political, and cultural characteristics of a city. For example, the websites of city governments in northern Germany and the Netherlands promote their own economic strengths and well-developed transportation networks, while the websites of remote cities in impoverished locations promote good natural ecosystems and long-lasting culture [27] (pp. 578–591). The postindustrial city of Rotterdam has experienced an industrial transformation from manufacturing and port transportation to an economy built on entertainment and leisure, commercial consumption, culture, and creativity. The pictures shown on the government website have also changed from blue-collar workers to white-collar/pink-collar workers and the creative class. Website marketing content exhibits “demasculinized” and “feminized” characteristics in line with “deindustrialization” and “tertiary industrialization” [28] (pp. 153–168). Stratford, a Canadian city with rich land resources, a good agricultural base, and a long industrial heritage, has a government website that focuses on images of local eco-agriculture, industrial tourism, museum tours, and performance art [29] (pp. 133–144).

However, due to the monopoly caused by the unidirectional network characteristics of Web 1.0, other studies have shown that the content of government websites may also be inconsistent with the city image. For example, Chicago’s Andersonville neighborhood was originally occupied by a Swedish immigrant community; although Swedish immigrants have all moved out of the city, Chicago continues to market Andersonville as a “Swedish immigrant community” with the aim of attracting multinational investment and skilled elites from Sweden [30]. The government website for Turin, Italy, promotes the city’s smart living, sustainability, gastronomic culture, luxury consumption, and the large amount of EU funding or project aid that the city has received in recent years to create an image of a “thriving economy and upgraded industrial structure”. In reality, however, Turin’s economy has been in steady decline since 2008, and it is heavily indebted [31] (pp. 1–7). The websites of the cities in the Randstad region of the Netherlands focus on good economic cooperation between the cities in the region, which is “highly networked with complementary industrial structures”; however, in fact, the cities of Amsterdam, Rotterdam, and Utrecht are relatively homogeneous in terms of industry, and there is more competition than cooperation [32] (pp. 2036–2056).

Inherent within the concept of Web 2.0 is the group of technologies—social media—that have facilitated the creation of a more socially connected Web, where anyone can participate by adding and editing content [33,34,35,36] (35, pp. 59–68; 36, pp. 168–182). More specifically, some scholars describe the social media driven Web 2.0 as “a collection of open-source, interactive and user-controlled online applications expanding the experience, knowledge, and market power of the users as participants in business and social processes...facilitating the flow of ideas and knowledge by allowing the efficient generation, dissemination, sharing, and editing/refining of informational content [37] (pp. 231–244).” The development of social media platforms has affected the method of communication between content producers and their audiences. Traditional media and early Web usage emphasize content delivery [38,39] (38, pp. 49–59; 39, pp. 41–58), but contemporary social media communication privileges and promotes interactions between users [40,41] (40, pp. 285–294; 41, pp. 357–365). Based on this new model, Ketter and Avraham (2012) and some other scholars identify four key elements in social media-based communications: (1) promote the creation of user-generated content, (2) leverage the online interaction into continuous relationships, (3) form virtual communities around a place to enhance conversation and interaction, and (4) learn from the customers, using the two-way interaction for feedback and market research [42]. This redefining of the relationship between the producer and consumer suggests a restructuring of power, where every participant can now be a producer of information. Traditional holders of power—such as municipalities—risk losing control of the messages that they produce through their branding and marketing, as they are unable to control the creation of information and discussion about their place.

In general, the attention given to user-generated content in online communities within the field of geography has focused on users’ geographic location information and the outcome characteristics of user-generated content while neglecting the exploration of the content production process [43] (pp. 683–695). Currently, mainstream social media, such as Twitter, TikTok, and Instagram, contain a large amount of pure marketing content, which exists only for the purpose of marketing objectives and does not reflect objective facts. Therefore, the cluttered nature of its information leads to a weakening of its ability to reflect the city’s image, which makes people subconsciously use the characteristics of the city as characterized by official information to validate the corresponding city image [44] (pp. 1939–1956). More importantly, the neglect of the production process of content in official media and online communities leads directly to the neglect of the groups behind them, and thus the relationship or “co-shaping” of social media, which presents a broader discourse, and official media, which is more local in nature, seems to be more useful than a single type of media datum for shaping and validating a city’s image.

As for the reason why we chose world cities in the United States, firstly, the world cities in the United States, such as New York, Los Angeles, and Chicago, occupy an important position in the global urban network. These cities are the core nodes of global finance, culture, transportation, and innovation, and can represent the image of developed-country cities in the process of globalization [45] (pp. 3–192). Studying these cities can help reveal the unique roles and development characteristics of developed countries’ world cities in the global system. Secondly, the United States has a large number of world cities and high data availability, which can cover as many cities as possible while respecting the regional characteristics of social media. In addition, the United States is the birthplace of social media, with platforms such as Twitter and Facebook having a wide user base that can reflect the public’s real-time perception and evaluation of cities. The rich official and social media data provide solid data support for comparing the similarities and differences in urban image between the two [46] (p. 3–14). Finally, the world cities in the United States have demonstrated diverse development paths and governance models. New York, as a global financial center, and Los Angeles, as a cultural and entertainment center, reflect the functional positioning of different cities in the global network. By analyzing these cities, we can delve into how different governance models affect the shaping and dissemination of urban image, especially in the contexts of accelerated globalization and digitization [47].

3. Data and Methods

3.1. Sample Cities

The sample cities were selected with reference to World Cities Ranking 2020 (https://www.lboro.ac.uk/microsites/geography/gawc/world2020t.html (accessed on 22 March 2023) published by the Globalization and World Cities Study Group and Network (GaWC), which classifies cities into five categories: Alpha, Beta, Gamma, High Sufficiency, and Sufficiency. This classification is constructed based on data, such as advanced producer services, and mainly expresses the degree of connectivity between a city and other cities on a global scale. The higher the classification, the stronger the connectivity. According to the latest World City Ranking 2020 report, 45 cities in the U.S. were ranked (

Table 1. U.S. world cities.

World City Class	Number	City
Alpha	5	New York, Los Angeles, Chicago, San Francisco, and Boston
Beta	13	Washington DC, Dallas, Miami, Houston, Atlanta, Denver, Philadelphia, Seattle, Detroit, Austin, Minneapolis, San Diego, and Tampa
Gamma	10	Charlotte, St Louis, Phoenix, Baltimore, Nashville, Cleveland, Kansas City, Milwaukee, Salt Lake City, and Columbus
High Sufficiency	5	Hartford, Raleigh, Indianapolis, San Antonio, and Cincinnati
Sufficiency	12	Jacksonville, Richmond, Oklahoma City, Des Moines, Louisville, Buffalo, Rochester, New Orleans, Omaha, Honolulu, Harrisburg, and Birmingham (Alabama)

).

3.2. Data Sources and Classification Index System

In this study, we collected data from both official and popular media. For the former, it was assumed that the homepage of a website is the focus of that website’s content, and since this study focused on textual content analysis, the text information on the homepage of official government websites was considered the target data. We drew on the content classification index system of government official websites established in previous research (

Table 2. Classification index system for textual content on government official websites.

Major Categories	Subclass	Weight
Management	Language category	0.2
	Government affairs	0.4
	Intercity connections	0.2
	Online office	0.2
Economy	Local business	0.4
	Foreign business	0.4
	Advertisement	0.2
Society	Job training	0.5
	Social services	0.5
Entertainment	Tourist services	0.25
	History introduction	0.25
	Travel services	0.25
	News updates	0.25

) [48] (pp. 86–99), consisting of 4 major categories—management, economy, society, and entertainment—as well as 13 subcategories, to collect and categorize the contents on the official government websites of the American world cities mentioned above. Due to the small number of data, manual identification was used to collect and count the official government website content data, which then had to be quantified. First, if the official government website of a city had a particular subcategory of website content, the corresponding subcategory of the city was assigned a value of 1; otherwise, it was assigned a value of 0. However, a different rule was applied in the case of the subcategory of language in the main category of management: when the official city government website had any language other than its national language, it was assigned a value of 1; if only its national language was present, it was assigned a value of 0. We thus obtained 45 × 13 binary quantization matrices for the United States cities.

In this study, we focused on analyzing the user-led generation of internet content on social media; thus, one of the most representative platforms, X, was chosen as the data source. As one of the most popular social media outlets in the United States, X is known for its real-time and topical discussions, and users tend to respond quickly to social events, news, and issues that have great significance for city image assessment. In contrast, some platforms may focus on sharing personal pictures or videos, with less immediacy and depth of discussion. At the same time, many politicians and social celebrities also express their views and participate in discussions on X, which can also give rise to highly discussed topics that are important for city image assessment. In addition, X offers primarily text-based content, allowing users to post brief opinions, comments, and discussions. Such textualized information can be more easily subjected to natural language processing and quantitative analysis to extract the sentiments and themes of public opinion. We crawled all content on X by using the names of each world city in the United States as keywords, covering a period of approximately 100 days from 1 November 2022 to 10 February 2023.

Since the contents of both social media and official government websites influence public opinion and behavior and have a very high degree of overlap in terms of the involvement of the urban marketing field, with drivers and stakeholders being in the same closed loop of social communication despite certain class differences, the basic views of drivers and stakeholders in the same field (e.g., urban marketing and urban image) present commonalities [48,49] (48, pp. 86–99). Therefore, we drew on the same content classification index system of government official websites established in the previous research mentioned above, which consists of four major categories (management, economy, society, and entertainment) [48] (pp. 86–99). In addition, as this study involved text classification in the field of natural language processing, we referred to the classification system of 20 Newsgroups, AG News, DBpedia, the British Broadcasting Corporation Corpus, Yahoo Answers, and Sogou News, which are commonly used for English text classification tasks. Based on this, we set up 21 subcategories for data classification [50,51,52] (50, pp. 1129–1153; 52, pp. 48–59). In the end, we constructed a classification indicator system for social media information consisting of 5 major categories (management, economy, society, entertainment, and advanced topics) and the abovementioned 21 subcategories (

Table 3. X textual content classification index system.

Major Categories	Subclass	Weight
Management	Office holder	0.5
	Politics	0.5
Economy	Finance	0.5
	Company	0.5
Society	Education	0.22
	Family and relationships	0.22
	Health	0.22
	Means of transportation	0.11
	Society and culture	0.11
	World	0.11
Entertainment	Entertainment and music	0.18
	Animal	0.09
	Artist	0.18
	Sports	0.18
	Film	0.09
	Nature	0.09
	Written work	0.09
	Building	0.09
Advanced topic	Computer	0.25
	Mathematics	0.25
	Technology	0.5

). In particular, the “advanced topics” category was created as a result of the analysis of the content of social media texts, which revealed the existence of a large number of opinions on basic sciences, cutting-edge technologies, and philosophical topics (opinions on current events, politics, life, worldviews, religions, etc.); such personal views and extended opinions, as well as assertions on future and irresponsible imaginings, could not be categorized into the first four categories corresponding to the official government website indicators. This classification index system was used to collect and categorize the internet textual content related to 45 American world cities on X. We crawled X text data with Python 3.11 programming and classified them by using a naive Bayes classifier [53]. To make the latter more target-specific, we adopted the six authoritative corpora mentioned above—20 Newsgroups, AG News, DBpedia, the British Broadcasting Corporation Corpus, Yahoo! Answers, and Sogou News—and used them to establish the index system for social media content classification for training, supplemented by the self-constructed corpus for X media content. The classification precision, recall, and F1-score were all above 0.9. Next, the documents in different categories of social media content per city were counted, and the proportion of the number of documents in each category to the overall volume of documents for that city was calculated, yielding a 45 × 21 matrix of document volume shares.

In this study, we weighted the two indicator systems above to make their indicator results more accurate and consistent with practical research applications. The weighting coefficient, λ, was determined by the relative importance of the subcategories to city development and public behavior. Since this study involved social and humanistic fields, the indicator had to correspond to the public’s views as much as possible; thus, the relative importance of each subcategory and tertiary classification depended on the corresponding interview results. We adopted the method of semi-structured interviews, in which the interviewee mentioned the subcategories of each major category and freely discussed the importance of each relatively to the major category. Considering that most social media users are young people, the interviewees were randomly selected as X users aged 20 to 35 years. In this study, we did not conduct expert interviews in each subcategory, only random interviews with ordinary online users to ensure outside–in “public” weighting. The weighting coefficients followed a rule: if a subcategory of a major category was significantly more important or mentioned more often than the others in the interview results, it was given a weighting coefficient that was an integer multiple of the weighting coefficients of the other subcategories.

It can be noted that the subcategories of the two indicator systems are different. This is due to the fact that, in this study, we focused on the differences between the groups communicating through the two data channels and did not compare their strengths and weaknesses in a specific aspect. And the academic community discussed the different characteristics of official data and social media data in informatics, and pointed out that these differences need to be considered when conducting data analysis and designing indicator systems based on their respective characteristics [54,55]. Therefore, the subcategories could not be limited to indicators that accommodated a single group but needed to be filtered according to the characteristics of the different media. Moreover, no matter which more detailed subcategories the media focused on, the public perception of the city image they contributed to was similar. For example, on the society front, the official websites focus more on job search, skills training, and basic livelihood guidance services, while the online community focuses on education, family relationships, healthcare, and transportation accessibility. Although different aspects are emphasized, both sources are centered on the most basic livelihood and social issues, giving the public the impression that “this city pays much attention to people’s livelihood”, and are thus categorized as “society”.

3.3. Data Standardization

Since the five major categories of the social media index included different numbers and weights of subcategories, we standardized the percentage share into the form of a Z score by using Z score normalization:

$$Z=\mathsf{\lambda }{\displaystyle \frac{X-\overline{X}}{\mathrm{s}}}$$

(1)

where $X$ represents the original data, $\overline{X}$ is the mean, $\mathrm{s}$ is the standard deviation, and $\mathsf{\lambda }$ is the weighting factor.

Next, the Z scores of all subcategories were weighted and summed, allowing for dimensionality reduction to obtain a matrix of Z scores for the 5 major categories of social media and one for the 4 major categories of official government websites. Since the range of Z scores involved negative numbers, the data interval was adjusted to [0.01, 1.01] by using the min–max normalization method for the convenience of subsequent calculations:

$$U_s={\displaystyle \frac{Z-\mathrm{m}\mathrm{i}\mathrm{n}\left(Z\right)}{\max \left(Z\right)-\mathrm{m}\mathrm{i}\mathrm{n}\left(Z\right)}}+0.01$$

(2)

Above, $U_s$ represents the normalized data, where s = 1 is the content index of the American official government websites and s = 2 is the content index of X. As a result, four types of matrices of American official government website text and social media text were obtained. The American social media matrices represent 5 different aspects of each world city that are highlighted on X, and those of the American official government websites represent 4 different aspects of each world city that are highlighted on the official government websites.

3.4. Cluster Analysis

We used systematic cluster analysis to further analyze the data. Then, combined with the one-way ANOVA value for the subclasses in the above index system, the differences among the various types of websites were calculated to choose an optimal clustering result. Finally, within-group linkages were used to calculate the distance between individuals and subclasses in the text matrix, and the furthest neighbor was used to calculate the distance between subclasses [56] (pp. 165–193). After determining the classification, we manually observed the data characteristics of each classification, summarized the distribution of dominant data for each type, and then confirmed its type characteristics based on the main category to which its dominant data belonged.

3.5. Relative Attention

In this study, in the context of the United States, we compared the textual content of official government websites with the corresponding social media content by using normalized and standardized data. We used the following formula to calculate the index ratio between the two types of content to derive the relative attention paid to each:

$${\rho }_{America}={\displaystyle \frac{U_1}{U_2}}$$

(3)

where $U_1$ is the content index of the American official government websites and $U_2$ is the content index of X. If 0 < ${\rho }_{America}$ ≤ 0.9, the attention paid to a particular subcategory on the official government websites is less than that on social media; if 0.9 < ${\rho }_{America}$ ≤ 1.1, the two aspects are comparable; if ${\rho }_{America}$ > 1.1, the attention paid to a subcategory on social media is less than that on the official government websites.

4. Results

4.1. Classification of Cities Based on Social Media and Government Websites

According to the clustering results, the considered American world cities were classified into five (A’, B’, C’, D’, and E’) and four categories (A, B, C, and D) according to the textual contents of the official government websites and the social media textual content, respectively.

4.1.1. Classification Based on Government Websites

With respect to the official government websites, we found that in Category A’ cities, attention is mainly paid to government affairs, online offices, local businesses, social services, and news and updates, i.e., internal economy and policy livelihood characteristics; in Category B’ cities, it is mainly paid to online offices, local businesses, social services, visitor services, and news and updates, i.e., internal economy and life entertainment characteristics; in Category C’ cities, it is mainly paid to government affairs, online offices, local businesses, job searches and training, visitor services, travel services, and news and updates, i.e., economic development, policy livelihood, and life entertainment characteristics; in Category D’ cities, it is mainly paid to multilanguage content, external business, job training, social services, visitor services, history, and news, i.e., external economy, policy livelihood, and life entertainment characteristics; finally, in Category E’ cities, it is mainly paid to multilanguage content and government affairs, i.e., external connections characteristics (

Table 4. Frequency and percentage of each type of American world city based on official government websites.

City Type	Category Characteristics	United States
		Cities	Frequency	Proportion
A’	Internal economy and policy livelihood	New York, Miami, Houston, Denver, Philadelphia, Tampa, Phoenix, Nashville, Milwaukee, San Antonio, Cincinnati, Harrisburg, and Birmingham (Alabama)	13	29%
B’	Internal economy and life entertainment	Chicago, Dallas, Austin, Cleveland, Salt Lake City, Columbus, Hartford, Raleigh, Richmond, Oklahoma City, Des Moines, Louisville, and Omaha	13	29%
C’	Economic development, policy livelihood, and life entertainment	Los Angeles, San Francisco, Boston, Washington DC, Detroit, San Diego, Charlotte, St Louis, Baltimore, Indianapolis, Jacksonville, Buffalo, Rochester, New Orleans, and Honolulu	15	33%
D’	External economy, policy livelihood, and life entertainment	Minneapolis	1	2%
E’	External links	Atlanta, Seattle, and Kansas City	3	7%

).

Overall, we found that the U.S. government believes that the homepage of official websites only needs to display a small amount of external information, while internal economic development, citizen life, social services, and happiness levels are more important and intuitive for urban marketing. Specifically, Categories D’ and E’ include world cities with an external nature, accounting for 2% and 7%, respectively, of the American world cities, totaling 9%. The content tendency of Category E’ cities is more monothematic, as only external connections are emphasized, with the objectives of communication and presentation, while D’ cities are represented as relatively slow-paced, with strong global economic potential, as they are more likely to seek economic cooperation and focus on people’s livelihoods. We found that government-level messages focus on inward economic development (including “economic development”) in Categories A’, B’, and C’, accounting for 29%, 29%, and 33%, respectively, a total of 91%, i.e., almost all of the U.S. world cities. In these cities, government communication focuses on the internal economy, as well as the lives and well-being of the citizens and social services. This shows that the United States is more inclined toward fostering internal development and stability. As the most-developed country in the world, it has the core strengths of urban attractiveness and global competitiveness and a well-developed media platform for exporting American culture and values; thus, for the government, official marketing has become less important, and internal services have become the mainstay of its functionality.

4.1.2. Classification Based on Social Media

In terms of social media, we found that, in Category A cities, public attention is mainly paid to real estate, stocks, society, and education, i.e., economic development and policy livelihood characteristics; in Category B cities, it is mainly paid to sports and popular entertainment, i.e., life entertainment characteristics; in Category C cities, it is mainly paid to social culture, sports, music, and entertainment, i.e., consumption enjoyment and policy livelihood characteristics; and in Category D cities, it is mainly paid to economy, science, and technology, i.e., high-tech economy and policy livelihood characteristics (

Table 5. Frequency and percentage of each type of U.S. world city based on social media.

City Type	Category Characteristics	Cities	Frequency	Proportion
A	Economic development and policy livelihood	Denver, Minneapolis, Milwaukee, Columbus, San Antonio, Jacksonville, Harrisburg, New York, San Francisco, Washington DC, and Oklahoma City	11	24%
B	Life entertainment	Los Angeles, Detroit, Cincinnati, and Louisville	4	9%
C	Consumption enjoyment and policy livelihood	Chicago, Miami, Houston, Atlanta, Philadelphia, Seattle, Austin, Tampa, St Louis, Phoenix, Baltimore, Cleveland, Kansas City, Indianapolis, Richmond, Des Moines, Buffalo, Rochester, New Orleans, and Birmingham (Alabama)	20	44%
D	High-tech economy and policy livelihood	Boston, Dallas, San Diego, Charlotte, Nashville, Salt Lake City, Hartford, Raleigh, Omaha, and Honolulu	10	22%

).

Overall, the American content sampled mainly focused on political issues, personal life, and social services. Specifically, nearly half of the U.S. world cities was in Category C (44%), thus characterized by a public that is interested in relatively high-consumption services or products, focuses on the enjoyment of life, and is attentive to the benefits and privileges that the government can provide them with. Category A and C cities are relatively contrasting in terms of public living standards, with A cities being perceived as experiencing fast-paced development but fostering an unsustainable pace of life and C cities as promoting a recreational and consumer-oriented attitude toward life. This is sufficient to indicate that, from the perspective of the public, the high-tech industry is closely related to the city and is likely to have numerous high-tech industries or, at least, the public attaches great importance to the high-tech industry. This may represent why the government has just issued some policies on the introduction of high technology. Finally, the proportion of United States Category B cities, which are characterized by a slow overall urban pace, and the relatively little attention is paid to their own economy and other areas of development (9%).

4.2. Relative Attention and Decisive Media Preferences

We used Equation (3) to calculate the relative attention paid to the four major categories for each of the considered United States cities and plotted the results into a line graph (Figure 1). In the graph, values of the vertical axis above 1.1 indicate that the attention paid to a particular aspect of the city on social media is less than that on the official government website, values below 0.9 indicate that the attention paid to a particular aspect on the official government city website is less than that on social media, and values between 0.9 and 1.1 indicate that a particular aspect is paid the same degree of attention on social media and official government websites; the latter case reflects the synchronization of internal guidance and external social media effects relative to the degree of attention paid to and the attitude toward the development of a certain aspect of the city. In addition, to improve the readability of the line graph, we removed extreme data from the dataset, including the following: 14.3, 10, and 70.8 for the management aspect of Houston, Tampa, and Cincinnati, respectively; 14.1, 33.7, 27.5, and 41 for the economic aspect of Los Angeles, Washington DC, Baltimore, and Oklahoma City, respectively; and 67 for Nashville’s social aspect. These data indicate that the official government websites of these cities pay much more attention to those aspects than social media users do, i.e., they are government-led aspects. Next, as shown in the figure, there are many world cities with data points very close to 0, indicating that the official government websites of these cities pay limited attention to the corresponding aspects, while social media users’ attention is relatively dominant.

Notably, for most cities, whether at the inside–out government level or the outside-in public level, there are active interest and involvement in city management online. In terms of economy, there are a considerable number of cities (11) in the United States where social media communication predominates, indicating that the economic aspects of these world cities are prone to be the focus of social media in today’s Web 2.0 environment. Moreover, most of the world cities in the United States that social media users pay attention to in terms of economic development are not top-tier cities, with their economic development level and world ranking being generally average. On the social front, only three cities in the U.S. have social media users attending to social issues, suggesting that there are few American citizens who need to express their social views and needs on social media. In terms of entertainment, there are some cities in the U.S. where related social media communication predominates, which is not surprising for this major category; entertainment is highly autonomous, and it is reasonable that the online public, as consumers of entertainment, pays greater attention to the entertainment aspect. At the same time, there is a positive association between entertainment and the consumption level, which suggests that there is greater consumption in the world cities with a dominance of social media attention paid to entertainment, which is confirmed by general perceptions in San Francisco, Houston, and Seattle.

5. Conclusions

In this study, we performed the first exploratory analysis of the correspondence, differences, and preference degree between city images shaped by internal and external information in the online environment by comparing local official government websites and online communities. By analyzing the turnover of the network era and the change in the power structure of urban discourse/information, we learned that information/content production and city marketing are shifting from unidirectional official platforms to more decentralized and externalized public platforms. At the same time, by combining correspondence and mutual validation of Web 1.0 and Web 2.0, we observed a complex relationship between local governments and online communities in terms of urban marketing and image shaping. Using U.S. cities as an example, we contribute to the literature on the media dimension of global cities. In response to the lack of theoretical research mentioned in the introduction, we present the following conclusions.

(1) Regarding the clustering analysis of city image, in the case of the considered U.S. world cities, we found that it can be identified and categorized based on Web 1.0 local official media content and Web 2.0 public online communities. By focusing on the internet content production process in typical information environments of different eras and combining programming crawlers, Bayesian text categorization, and cluster analysis, we accurately categorized online city images according to various city aspects. The results on the shaping of city image were obtained by analyzing official city image marketing channels, where information channels are unidirectionally monopolized from the inside–out, and decentralized external information channels. On the one hand, the era of globalization has increased the need for city image brand marketing in most cities. As a result, local official media (official government websites) have become overly “economically decorated”, which refers not only to the city’s economic level, but also to the intention of international cooperation, the level of infrastructure, industrial policy, and other economically oriented marketing representations. On the other hand, the boom in social media has led to a change in the structure of public access to and the production of information. There is now a dichotomous dialog between “inside–out and outside–in” image shaping in terms of the people communicating through official and social online platforms. The shift in the monopoly of discourse and information has led to atypical changes in the city image, with the most notable difference being that the entertainment and livelihood characteristics depicted by social media have become more important for characterizing a city than the image portrayed by the local official media, an aspect that seems to have been overlooked in the local government long-standing marketing image of city economic development, even if we can perceive such characteristics quite clearly. As we no longer look to official information channels as the only measure of city image, it seems that more complex and rich city personalities are gradually revealing themselves to us. Compared with local government monopoly marketing, in the current Web 2.0 network era, national or even global public comments or assertions are more influential [57] (pp. 1–16), especially in the construction and identification of city image. While the typical stereotypes of cities were developed in the global competition of the Web 1.0 era, in today’s booming social media information clusters, such stereotypes have evolved into an effective base image [58] (pp. 123–158). This base image is mostly used by outsiders to validate urban outcomes, as well as geographic imagery with migratory tendencies (e.g., having a detailed knowledge of many aspects of the city because of plans to relocate to the city in the future) [59,60]. However, social media also has some limitations in reflecting the image of a city due to the specificity of the motivation for content production, and it remains an unresolved question as to whether or not Web 2.0 can truly replace Web 1.0 in its entirety.

(2) To determine the significance that the differences between the groups communicating through official and social media channels have in shaping the city image, in this study, we compared the logics and motivations of these two groups. The evidence from our preliminary analysis suggests that, although the Web 1.0 era is viewed as past or obsolete online [6,7,8,9] (6, pp. 181–198; 7, pp. 211–225; 8, pp. 233–259), its importance in shaping the city image should not be underestimated. For instance, for the public outside the city, to “validate” the city’s image, the information on Web 1.0 channels still has the ability to echo social media in the Web 2.0 era and can play a “corrective” role with respect to the bias of social media image building due to the motivation for content production. Although both are considered to be “updates” of the online environment on the timeline [12] (pp. 93–106), by analyzing their content production processes, we found that they form two information environments with different subjects, needs, and marketing outcomes, and that the people or subjects communicating through these channels have a certain degree of internal–external duality. In the new era, the accurate judgment and classification criteria of city image should be comprehensive [61,62,63,64] (61, pp. 1793–1809; 62, pp. 1283–1299; 63, pp. 1791–1815; 64, pp. 1923–1946); from a media perspective, this is manifested through the internal and external interactions of the groups communicating through the different types of media.

(3) Finally, while much of the research discussion in the field of city image has been about the possibility of generating urban marketing models that are broadly adaptable [65] (pp. 1–10), in this study, we developed a city-specific research proposal. While the abovementioned points have been shown to contribute to a more integrated approach to global city branding in American world cities, the duality of “local” and “ external” shaping is crucial to the city image and to the geographic imagination of outsiders in specific contexts [66]. Preference confirmation between the two can lead to more accurate city image identification in the Web 2.0 era. In this study, we investigated the preferences for the semantically opposed notions of “inside–out” and “outside–in” and found prima facie evidence of the existence of widely differing inclinations. Preference identification would allow for easily extracting one or more cities as representative and specific cases from a broadly adapted urban marketing model, representing a potential literature contribution to the study of individual urban experiences in human geography.

6. Discussion

Based on the three conclusions above, it can be recognized that urban image can be divided into several typical categories through different types of media, and the specific urban image of each city can be quantitatively subdivided through the preference level of different media, thus obtaining a unique urban image. This process is reversible, and after developing accurate quantitative identification methods, there will naturally be specific quantitative shaping methods for it. In addition, the research sample of this article, world cities, as important cases for studying globalization and global urban development, can to some extent enhance the global universality of the conclusions [67]. Therefore, it can be said that each city (taking the world cities in the United States as an example) has its own unique image compared to other cities, but there are some similarities or subtle differences between them. In the past, city images could only be broadly classified. However, this article provides a method by which each city can shape or be recognized for its unique image through the concepts of “inside–out” and “outside–in” dualism, as well as the weight basis provided by the degree of preference between the two. Although the difference between these images may only be an extremely small quantitative difference, when it comes to different aspects of each city, they will converge into a multidimensional and multi-level image difference.

The main theoretical contribution of this article is that, firstly, in the field of human geography, traditional research on urban image mainly relies on the analysis of the physical characteristics and historical and cultural background of geographic space (such as Kevin Lynch’s “Urban Imagery”). However, with the development of the internet, especially the social media in the Web 2.0 era, the construction of city image has expanded from physical space to virtual space. This study indicates that, under the dual influence of Web 1.0 and Web 2.0 information environments, the construction process of urban image exhibits a shift from unidirectional to multidirectional interaction. This conclusion reveals that the diversified urban image in virtual space goes beyond the traditional geographical research’s focus on physical space, filling the theoretical gap in how human geography understands and analyzes the process of constructing urban image in virtual information space. The “spatial production” theory in traditional human geography (such as Henri Lefebvre’s theory) mainly focuses on the social production process of physical space, neglecting the role of virtual information space in shaping modern urban image, especially emerging social media content [68,69]. This study addresses this limitation by proposing that, in virtual spaces, the image of cities is also influenced by changes in the power structure of information production. Secondly, in the field of urban marketing and brand building, classic theories such as Kotler’s “urban marketing” model emphasize market-oriented promotion through the development of strategies and the utilization of the unique characteristics of the city [70]. However, this model mostly involves top-down unidirectional brand building, emphasizing government or official agency led information transmission. This study found that social media plays an increasingly important role in shaping urban brands in the Web 2.0 environment, forming a two-way interaction between official and public information. This discovery indicates that social media is not only a carrier of information dissemination in urban brand evaluation, but also a participant in brand image shaping. Finally, the contribution to the theory of world cities, such as Sassen’s “Global Cities” theory, focuses on the central position of nodal cities in the global economic system in finance, commerce, culture, and other aspects [71]. Traditional research often focuses on hard power indicators, such as physical infrastructure and the distribution of multinational corporations’ headquarters, while neglecting soft power, especially the impact of online information flow on urban status. This study reveals that, in the context of globalization and the rise of social media, the shaping of local image has shifted from traditional economic hard power to online soft power competition by analyzing the city image shaping on official government websites and social media.

The practical significance of this study is: Firstly, the dual information environment analysis model proposed in this article can help urban managers better understand the information exchange and feedback mechanism between the public and the government. This mechanism can identify information asymmetry: by comparing and analyzing official statistical information and social media discussions, urban managers can discover which areas have information asymmetry (such as the gap between the government’s focus on promoting economic development and the public’s actual attention to social welfare areas). This analysis helps identify blind spots in policy implementation and adjust public policies in a timely manner, making policy formulation and implementation more in line with public needs. Secondly, it is possible to optimize urban communication strategies, and research has revealed the different roles of Web 1.0 and Web 2.0 in the construction of urban image, which has practical guidance significance for governments and official institutions to formulate more effective information dissemination strategies. Finally, and most importantly, it has practical significance to enhance the competitiveness of urban brands. This article reveals the cognitive differences in urban brand image between different audiences (such as local and external groups) by comparing official information and social media discussions. This helps city managers adopt differentiated marketing strategies in brand promotion, adjust promotional content for different audiences, and enhance the diversity and inclusiveness of brand image. In addition, this article provides a city image evaluation method based on social media big data, which can reflect the public’s evaluation and feedback on the city brand in real time, help city managers quickly respond to market changes, enhance the overall competitiveness of the city, and strengthen its global network position and competitive advantage.

As for the limitations of this article and the potential for future research, firstly, this article mainly relies on text data from official government websites (Web 1.0) and social media platforms (Web 2.0). These data may have selective bias: on the one hand, the text content of official websites tends to carry functional information, while images and videos may to some extent better reflect the content of urban landscape and imagery. On the other hand, the discussion content on social media may be influenced by the specific preferences of platform user groups, hot event drivers, and algorithm recommendation mechanisms, making it difficult to fully represent the views of the entire society. Especially when using data from specific social media platforms, inactive groups or user groups using other social platforms may be overlooked. In this regard, future research can introduce multi-platform and multi type data (such as Little Red Book, TikTok, Facebook, etc.; pictures and videos, etc.), and combine offline surveys or questionnaires to enhance the universality and representativeness of data. Future research can integrate different data sources and types, use multi-source data fusion technology, reduce data bias, and obtain a more comprehensive urban image and public perception. Secondly, this article mainly analyzes static data from a certain period of time, but the urban image and public attention points present dynamic changes over time. Especially in the event of emergencies (such as natural disasters, policy changes, and major holidays) or specific social backgrounds (such as economic downturns and changes in international cooperation), the image of a city may undergo rapid and significant changes. In this regard, future research can adopt time-series analysis or dynamic tracking methods to collect data over a long-time span, in order to observe and analyze the evolution process of urban image at different time points. Finally, this article ignores the emotional tendencies of social media content. Although it analyzes the concerns discussed by the public, it does not measure the positive and negative aspects of comments, which may result in some deviations in the final image determination. Future research should consider developing more refined urban image assessment models, constructing multi-level and multi-judgment comprehensive evaluation indicators, and combining various analysis techniques, such as sentiment analysis, theme modeling, and scenario analysis, to provide a comprehensive understanding of urban image.