GEO: Lies and Truth—Should Small Businesses Invest in AI Search Visibility?
Small businesses are increasingly being offered services that promise greater visibility and more recommendations in ChatGPT, Gemini, Perplexity, and other AI platforms.
The difficult question is not whether AI-driven discovery matters. It is which GEO services are worth paying for—and which claims should be ignored.
Generative Engine Optimization, or GEO, refers to efforts intended to increase how often a company, product, or website is mentioned, cited, or recommended in generative AI answers.
Providers may recommend tactics such as:
Writing in a style supposedly preferred by AI
Adding large numbers of questions, answers, lists, and tables
Expanding structured data
Installing an
llms.txtfileCreating AI-specific Markdown or summary pages
Spreading brand mentions across Reddit, Quora, and other platforms
Measuring a brand’s share of AI answers
Managing its position in ChatGPT recommendations
But do these methods work?
Businesses should measure how AI services describe and recommend them. However, many tactics sold as proprietary GEO techniques remain unproven, provide benefits already associated with conventional SEO, or promise outcomes that no outside provider can control.
Small businesses should not begin with an expensive GEO platform or consulting contract.
They should first make sure that their website can be discovered, their official business information is accurate, their content answers real customer questions, and their reputation is supported by genuine customer experiences and credible external sources.
This article evaluates major GEO claims using five categories:
Fact: Supported by official documentation or sufficient evidence
Partially true: Provides a legitimate benefit, but that benefit is exaggerated when presented as a special GEO advantage
Insufficient evidence: Possible in theory, but its effectiveness has not been demonstrated
Cannot be guaranteed: An outcome that no outside provider can control or promise
Avoid: A tactic that lacks proven value and may create policy or reputational risks
Is GEO an entirely new marketing discipline?
Verdict: Partially true. AI-generated answers are a new customer touchpoint, but most GEO practices extend existing SEO, business information management, content strategy, and reputation management.
The problem that GEO attempts to address is real.
AI services may use websites, search results, business profiles, reviews, news coverage, community discussions, and other sources when generating answers.
If a company’s online information is inaccurate or incomplete, an AI service may describe the business incorrectly, cite an outdated source, or exclude it from the options presented to a customer.
However, most of the work required to address these problems existed before GEO:
Managing website crawling and indexing
Maintaining accurate and consistent business information
Clearly explaining products and services
Publishing content that answers customer questions
Managing reviews and external reputation
Earning coverage from credible third-party sources
Tracking visits, inquiries, bookings, and purchases
The genuinely new responsibility is monitoring how a brand appears in AI-generated answers, identifying inaccurate descriptions, examining the sources used in those answers, and tracking whether AI discovery leads to meaningful business results.
Google now directly addresses GEO and AEO in its official guidance. It states that its generative AI features rely on existing Search ranking and quality systems and that, from Google Search’s perspective, optimizing for generative AI search remains part of SEO. Google’s guide to optimizing for generative AI features
This guidance applies specifically to Google Search. It does not prove that ChatGPT, Gemini, Perplexity, and every other AI service use identical systems.
GEO is best understood as an extension of SEO, Business Discovery, content strategy, and reputation management into AI-generated answers—not as a replacement for those disciplines.
Is there a special writing style that AI prefers?
Verdict: Partially true. Clear and readable writing matters, but there is no proven writing formula that AI systems universally prefer.
Some GEO providers recommend dividing content into short sentences and paragraphs and adding large numbers of questions, answers, lists, summaries, and tables.
These formats can be useful.
A direct answer can help a reader find important information quickly. A table may clarify a genuine comparison. A list can make a set of conditions or steps easier to understand.
However, these are established principles of effective web writing. They are not exclusive GEO techniques.
There is no published rule demonstrating that a particular sentence length, paragraph structure, question format, or keyword arrangement will reliably increase citations or recommendations across generative AI services.
Google states that website owners do not need to break content into tiny sections or rewrite it in a special way for generative AI search. Its systems can understand synonyms, general meaning, and relevant passages within a page. Google’s guide to optimizing for generative AI features
That does not mean structure is irrelevant. Poorly organized, unclear, or incomplete content can still be difficult for both customers and information systems to use.
The distinction is important:
Clear structure is useful.
Artificially fragmenting content for AI is not a proven advantage.
Direct answers are useful when customers need them.
Rewriting every page into a supposed AI formula is unnecessary.
Lists and tables should improve understanding, not simply increase their frequency.
Content should answer genuine customer questions accurately and clearly. Small businesses do not need to pay for a separate service merely to apply an “AI-friendly writing style.”
Will publishing more questions and answers increase AI citations?
Verdict: Insufficient evidence. Answering relevant customer questions is useful, but increasing the number of FAQs does not automatically increase AI citations or recommendations.
FAQs can provide real customer value when they address recurring questions about:
Pricing
Eligibility or service conditions
Booking procedures
Service areas
Shipping
Returns and refunds
Installation
Maintenance
Warranties
The problem begins when similar questions are mass-produced solely to target AI systems.
A company might create separate pages for multiple versions of the same question or automatically generate hundreds of answers that customers rarely ask.
This can increase the amount of content without increasing its usefulness.
It also creates an operational problem. If prices, policies, services, or business hours change, some answers may be updated while others remain outdated. The website may then contain conflicting versions of the same information.
Google warns against creating separate content for every possible search variation when the primary purpose is manipulating rankings or generative AI answers. At sufficient scale, such activity may conflict with its scaled content abuse policy. Google’s guide to optimizing for generative AI features
This does not mean businesses should avoid FAQs. It means they should choose questions based on actual customer needs rather than an assumed AI preference.
Useful sources for identifying questions include:
Customer inquiries
Sales conversations
Support requests
Search data
On-site search records
Reviews
Objections raised before purchase
Questions asked during booking or consultation
FAQs should be created when they resolve genuine customer uncertainty. There is no reason to pay for mass FAQ production whose only purpose is increasing AI citations.
Does structured data improve AI recommendations?
Verdict: Partially true. Structured data can help search engines interpret page information and determine eligibility for rich results, but it is not a separate technology for securing AI recommendations.
Structured data presents information about products, prices, businesses, events, reviews, and other entities in a standardized machine-readable format.
When implemented correctly, it may help search systems understand certain information and determine whether a page is eligible for specific search features or rich results. The structured data must also match the content visible on the page. Google’s introduction to structured data
Relevant structured data may therefore be useful for ordinary website and search management.
For example, a business may use appropriate markup for:
Local business details
Products
Offers
Events
Articles
Breadcrumbs
Videos
However, there is no evidence that adding structured data alone will make ChatGPT or another generative AI service recommend a business more often.
Google explicitly states that structured data is not required for its generative AI search features and that there is no special Schema.org markup that businesses need to add for them. Google still recommends using appropriate structured data as part of an overall SEO strategy because it can support eligibility for rich results. Google’s guide to optimizing for generative AI features
Again, that statement describes Google Search. Other services may process structured information differently, but no general recommendation guarantee follows from adding markup.
If a website lacks relevant structured data, the business can implement it as part of ordinary SEO. It should not pay a separate premium for structured data presented as a proprietary GEO method that guarantees AI recommendations.
Is llms.txt necessary for AI visibility?
Verdict: Insufficient evidence. There is no proof that llms.txt increases brand mentions, citations, or recommendations across major AI services.
llms.txt is a proposed standard for presenting important website information and links in a format intended to be convenient for large language models.
The proposal was introduced in September 2024. Its existence shows that some developers believe a standardized file could help AI systems locate useful resources. llms.txt proposal
However, proposing a standard does not demonstrate that the standard improves visibility.
Google states that Google Search does not use llms.txt and that businesses do not need the file to appear in either conventional or generative AI search. Google further explains that the file neither helps nor harms visibility or rankings in Google Search because Google Search ignores it. Google’s guide to optimizing for generative AI features
This conclusion should not be expanded beyond the available evidence.
Other services or systems may choose to use llms.txt now or in the future. But there is currently insufficient evidence that installing the file will increase citations or recommendations in ChatGPT or across major AI platforms.
The practical decision depends on cost.
If a technically capable business can add and maintain the file with little effort, doing so may be a low-risk experiment. It should not distract from more important problems such as inaccessible pages, inaccurate information, thin content, or broken customer links.
llms.txt should not currently be treated as a priority. If a provider presents it as a core GEO technique or charges a substantial fee to install it, the investment is not justified by the available evidence.
Will allowing AI crawlers increase recommendations?
Verdict: Access is a genuine prerequisite, but allowing crawlers is not an optimization technique that increases recommendations.
For a web page to be searched, summarized, or cited through a search-based AI system, the relevant service generally needs to be able to access it.
OpenAI advises publishers not to block OAI-SearchBot if they want their content to be eligible for inclusion in ChatGPT search summaries and citations. Businesses can inspect their robots.txt settings to determine whether access is being blocked. OpenAI Publishers and Developers FAQ
Allowing access, however, does not guarantee inclusion.
It only removes one possible technical barrier. It does not resolve problems involving:
Relevance to the customer’s question
Information accuracy
Content quality
Source credibility
Website authority
Conflicting external information
The AI service’s selection and answer-generation process
Crawler policies also vary by service and purpose. A crawler used for search indexing may not be the same as one used for model training or other functions. Businesses should review the current documentation for each service rather than treating all AI crawlers as interchangeable.
Businesses should verify that relevant search crawlers are not unintentionally blocked. There is no reason to purchase an expensive GEO service solely to make this basic technical adjustment.
Do more external mentions lead to more AI recommendations?
Verdict: Partially true. Genuine reviews and independent external reputation matter, but there is insufficient evidence that artificially increasing brand mentions will cause AI systems to recommend a business.
AI services do not necessarily rely only on a company’s official website.
Depending on the service and question, they may also use search results, review platforms, news coverage, industry websites, videos, forums, community discussions, and other external sources.
Genuine customer reviews, independent assessments, expert commentary, and credible industry coverage can therefore help customers and information systems understand a business.
The problem begins when companies attempt to manipulate this principle by planting artificial mentions. This practice is sometimes sold as “LLM seeding.”
Examples include:
Posting self-promotional recommendations on Reddit or Quora
Using paid or controlled accounts to imitate customer experiences
Registering a company in large numbers of irrelevant directories
Republishing similar promotional content across multiple websites
Adding promotional business information to Wikipedia
Paying for undisclosed recommendations
Repeating brand names solely in the hope that AI systems will absorb them
No provider can guarantee that these activities will cause a particular AI service to mention or recommend the brand.
Google also advises against seeking inauthentic mentions for generative AI visibility. Its official guidance explains that its core ranking systems focus on high-quality content while other systems detect and block spam. Google’s guide to optimizing for generative AI features
The value of external reputation depends on authenticity, relevance, independence, and credibility—not merely the number of times a brand name appears.
LLM seeding based on fake experiences, disguised self-recommendations, irrelevant directory submissions, or duplicated promotional content should be avoided. Its effect on AI recommendations is unproven, and it may violate platform policies or damage the brand’s reputation.
External reputation should be earned through genuine customer experiences and independent evaluation, not artificially planted.
Should businesses create separate AI-only pages?
Verdict: Insufficient evidence. There is no proof that duplicating existing content in an AI-specific format increases citations or recommendations.
Some GEO providers recommend creating separate Markdown pages, AI summaries, or large collections of question-based pages that supposedly make information easier for AI systems to process.
A separate page can be useful when it serves a distinct customer need. For example, a detailed comparison, technical documentation page, pricing explanation, or service eligibility guide may deserve its own page.
The problem is not the file format or the existence of a new page. It is the absence of additional value.
Repeating essentially the same content in another format does not mean that AI services will cite or recommend the business more often.
Duplicated pages also create more information to maintain. When a price, service, policy, location, or operating hour changes, one version may be updated while another remains outdated.
Google states that businesses do not need to create special machine-readable files, Markdown pages, or AI-specific versions of their content to appear in Google Search’s generative AI features. It also recommends reducing unnecessary duplicate content. Google’s guide to optimizing for generative AI features
This does not establish how every AI service processes every format. It does show that AI-only duplication is not a general requirement for Google Search visibility.
Before creating AI-only pages, businesses should make their existing pages accurate, useful, accessible, and easy to maintain. New pages should be created for genuine customer value—not solely for AI systems.
Do GEO metrics show real business performance?
Verdict: Useful within limits. GEO metrics can help businesses observe selected AI answers, but they do not measure total customer exposure, actual market share, or revenue performance.
GEO tools may provide metrics such as:
Mention Rate
Citation Rate
Recommendation Rate
AI Share of Voice
Visibility compared with competitors
Answer accuracy
Citation source share
These metrics can help a business examine how it and its competitors appear in answers to a defined set of questions.
“AI Share of Voice” should be understood as a selected-query visibility metric. It usually represents how often a brand appears within the questions and tests chosen by the measurement provider. It is not the brand’s share of the entire AI market or its share of actual customer demand.
The limitations are significant.
AI services generally do not disclose complete usage volumes for individual questions. It may therefore be impossible to know how often real customers ask the exact questions selected by a GEO tool.
Answers can also vary according to:
Question wording
Location
Language
Account status
Personalization
Previous conversation context
Search date
Retrieved sources
The model or answer system being used
A brand appearing in one response does not mean that it has secured consistent visibility.
Mentions, recommendations, and citations must also be separated.
A brand may be mentioned without being recommended. Its website may be cited even when its product is not selected. A business may be recommended without receiving a website visit, inquiry, booking, or purchase.
Measurement is improving, particularly within Google’s own environment.
Google Search Console now provides dedicated Generative AI performance reports covering impressions within AI Overviews, AI Mode, and generative AI features in Discover. This offers more reliable first-party measurement of visibility within Google’s generative AI search experiences. However, it still does not disclose demand or usage volume for every individual customer question. Google’s announcement of Generative AI performance reports
Third-party GEO tools can still be useful for repeatable question testing and cross-platform monitoring. Their results must be interpreted within the conditions under which the tests were conducted.
A claim such as “AI Share of Voice increased by 300%” should not be accepted without context.
An increase from 1% to 4% can be presented as 300% growth. If the provider does not disclose the question set, number of repeated tests, measurement period, location, language, account conditions, and comparison method, the figure says little about actual performance.
Businesses should not invest in a service merely because it increases an AI mention metric. GEO tools become more valuable when their measurements can be connected to visits, inquiries, bookings, purchases, or another meaningful business outcome.
Can anyone guarantee higher rankings in ChatGPT?
Verdict: Cannot be guaranteed. No outside provider controls which businesses ChatGPT or another AI service will recommend for a particular question.
AI providers disclose some information about how their search and answer systems work.
However, they do not publish the complete criteria and weighting used to decide which businesses will be mentioned, cited, or recommended in every situation.
Answers may change according to:
The wording of the question
The user’s location and language
The time of the search
Account settings and personalization
Previous conversation context
The web pages and external sources retrieved
The answer-generation system used by each AI service
The following promises therefore attempt to sell outcomes that cannot be guaranteed:
Guaranteed top visibility in ChatGPT
Guaranteed AI recommendation rankings
A promise that AI will learn a brand within a few weeks
Guaranteed recommendations after completing a particular task
Number-one visibility across all major AI services
Improving a website, business information, content, and external reputation may increase the likelihood that an AI service will understand or describe the brand accurately.
That is different from controlling the answer.
Businesses should not sign contracts with providers that promise top ChatGPT visibility, number-one AI recommendations, or guaranteed brand learning within a specified period.
What should businesses actually do about GEO?
The practical value of GEO does not come from secret files, rigid writing formulas, or techniques intended to manipulate AI systems.
Businesses should separate the genuinely new work introduced by AI-generated answers from established work that remains essential.
Work that is genuinely new to GEO
Measuring brand visibility for questions that real customers may ask
Distinguishing simple mentions, inclusion among options, active recommendations, and source citations
Checking whether AI services describe the business and its products accurately
Analyzing why competitors are included and which sources support them
Identifying external sources used in AI-generated answers
Tracking visits and conversions originating from AI services
Existing work that has become more important
Resolving website crawling and indexing problems
Maintaining accurate official business and product information
Managing information consistency across platforms
Publishing content that answers customer comparison and selection questions
Providing original expertise and evidence based on real experience
Managing reviews and independent external trust signals
Checking broken booking, inquiry, and purchase links
Monitoring and correcting changed information over time
The main new responsibility introduced by GEO is measuring and managing AI-generated answers as another customer discovery touchpoint. Most other activities remain part of established SEO and Business Discovery operations.
How should small businesses test GEO?
Small businesses do not need to ignore GEO.
They should, however, begin with a limited test rather than an expensive platform or long-term consulting contract.
Start by selecting 20 to 30 questions that customers might genuinely ask when comparing or choosing products and services.
Do not use only questions containing the company’s brand name. The set should reflect actual customer problems, conditions, comparisons, and decisions.
For example:
Which businesses provide this service in my area?
Which product is suitable for a particular situation?
What is the difference between A and B?
What should I check before choosing this service?
How much does it cost, and who is it suitable for?
Which provider meets a particular set of requirements?
Choose two or three AI services relevant to the business and record the initial results.
Evaluate each answer using separate criteria:
Was the brand simply mentioned?
Was it presented as one option among several?
Was it actively recommended?
Was the official website cited?
Was the information accurate?
Why were competitors included?
Which external sources were used?
Then correct problems that can be clearly verified:
Outdated business hours
Incorrect service descriptions
Inconsistent business names, addresses, or phone numbers
Broken booking and inquiry links
Important pages missing from search indexes
Content that fails to answer essential customer questions
Repeated inaccurate information from external sources
After making the changes, test the same questions under the same conditions for 8 to 12 weeks.
Whenever possible, keep the following conditions consistent:
AI service and model
Question wording
Location
Language
Account or non-account status
Testing frequency
Number of repeated tests
Do not evaluate only mentions and recommendations. Track business outcomes as well:
Visits from ChatGPT and other AI services
Google’s generative AI impressions
Brand search volume
Direct website traffic
Customer-reported discovery sources
Inquiries
Bookings
Purchases
OpenAI states that referral URLs from ChatGPT search automatically include utm_source=chatgpt.com. Businesses can use this parameter in analytics tools to identify inbound traffic from ChatGPT search results. OpenAI Publishers and Developers FAQ
Google’s Generative AI performance reports can provide first-party visibility data for Google’s AI Overviews, AI Mode, and generative AI features in Discover. Google’s announcement of Generative AI performance reports
Businesses should expand their use of professional GEO tools or services only after the test produces meaningful evidence of improved discovery, visits, inquiries, bookings, or purchases. If the change cannot be connected to business performance, they should not increase their investment.
What should businesses ask a GEO provider?
When evaluating a GEO service, businesses should examine how the provider measures and verifies performance—not how many features the platform offers.
Ask the following questions:
Which AI services are measured?
How are questions that real customers ask identified?
How is demand for each question estimated?
Are mentions, recommendations, and citations measured separately?
How many times is the same question repeated?
How are differences in location, language, account, and personalization handled?
Under what conditions are results before and after GEO work compared?
How are the effects of GEO separated from SEO, content, PR, and advertising?
Are visits, inquiries, bookings, and purchases from AI services tracked?
Can the service be discontinued if no meaningful effect is observed?
Does the service include paid recommendations or artificial external mentions?
Is there independent evidence that
llms.txtor AI-specific pages improve performance?What evidence supports any promise of guaranteed AI recommendations?
The provider should also disclose:
The full or representative question set
The number of tests
The testing period
The services and models used
The locations and languages used
The definition of each metric
The baseline and comparison method
The difference between visibility metrics and business outcomes
If a provider refuses to disclose its questions and measurement conditions—or cannot explain how AI visibility connects to business results—its reported growth rates should not be trusted.
Businesses should not contract with providers that guarantee recommendation rankings or claim that they can make AI systems learn a brand.
Capsule Rating
The subject being evaluated is a limited test in which a small business measures its current visibility in AI-generated answers and corrects basic, verifiable problems.
| Criterion | Rating | Reason |
|---|---|---|
| Impact | 3/5 | AI Discovery is becoming more influential, but its actual impact on customer acquisition has not been established for most small businesses. |
| Urgency | 3/5 | Businesses should assess their current position and establish baseline data, but full-scale investment is not yet urgent. |
| Business Fit | 3/5 | It is particularly relevant to businesses that depend on comparison, recommendation, local discovery, or professional services, but it does not apply equally to every small business. |
| Cost to Respond | 2/5 | Selecting questions, checking essential information, and conducting limited monitoring can begin with existing staff and little or no additional budget. |
| Evidence Confidence | Medium | The need to monitor AI-generated answers is credible, but evidence that individual GEO tactics increase recommendations remains limited. |
Recommendation: TEST
This does not mean testing expensive GEO technology.
Select 20 to 30 genuine customer questions and two or three relevant AI services. Document the current results, correct clear problems in the existing website and business information, and observe changes for 8 to 12 weeks.
Expand the work only if it produces measurable improvement in customer discovery or business outcomes.
At present, small businesses should not prioritize paid implementations of llms.txt, AI-specific writing formulas, mass-produced FAQs, duplicated AI-only pages, artificial LLM seeding, or services that guarantee AI recommendations.
Key Takeaway
GEO introduces one important new responsibility: businesses now need to measure and manage how they appear in AI-generated answers.
That does not make established SEO and Business Discovery work obsolete. Accurate information, accessible websites, useful original content, genuine reputation, and continuous monitoring remain the foundation.
Small businesses should test AI visibility within a limited and measurable scope. They should increase their investment only when improved AI visibility can be connected to visits, inquiries, bookings, purchases, or another meaningful business result.
댓글
댓글 쓰기