GEO for Small Businesses: How a Lesser-Known Brand Gets Cited by ChatGPT and Google (2026 Data)
Updated November 4, 2026. Every figure links to its original source. Study limitations (perception surveys, small samples) are flagged in the text and in the Methodology section.
Key takeaways:
Generative Engine Optimization (GEO) is about getting a brand mentioned and cited in the answers of ChatGPT, Gemini, Perplexity, Copilot and Google's AI Overviews, not just ranked in a list of links.
In an NP Digital survey of 812 respondents (October 2026), 56% named low brand recognition (32%) or a lack of independent mentions (24%) as the single biggest obstacle for smaller brands.
Measured data points the same way: more than 84% of the links AI tools cite come from earned media or third-party sources, versus 15.2% from companies' own sites (Muck Rack, October 2026).
The opening is specificity. 63% of respondents feel confident they can answer a question about a customer's specific situation better than a larger competitor. For a broad category question, that drops to 27%.
Google states it plainly: no special markup, no "AI" file and no particular schema is required to appear in AI Overviews.
"Why does ChatGPT always recommend the same big brands?" For a small business, it's a fair question. The short answer: language models rely heavily on what the rest of the web says about you. The useful answer: there is ground where a specialized company starts ahead, and the data published in 2026 maps it out quite precisely.
This guide pulls that data together, checked against each original publication, separates what has been measured from what marketers believe, and turns it into an action plan.
What is Generative Engine Optimization (GEO)?
GEO covers the editorial, technical and PR practices that increase the odds a brand will be mentioned, described accurately and cited as a source in an AI-generated answer.
The term comes from a research paper by Pranjal Aggarwal and colleagues, presented at KDD 2024. The authors tested different ways of reworking a text and report that these methods can boost a source's visibility by up to 40% in generative engine responses (arXiv, 2024).
Traditional SEO | GEO | |
|---|---|---|
Goal | Earn a position on the results page | Get mentioned or cited in the answer |
What the searcher sees | A list of links | A summary with a handful of sources |
What you measure | Rankings, impressions, clicks | Presence in answers, citations, accuracy of the description |
Signals that carry weight | Relevance, technical quality, backlinks | Relevance, brand mentions across the web, verifiable evidence |
GEO doesn't replace SEO. To be cited in AI Overviews or AI Mode, a page first has to be indexed by Google and eligible to show with a snippet (Google Search Central). But ranking well is no longer enough: according to Ahrefs, only 37.9% of the pages cited in AI Overviews rank in Google's top 10 (Ahrefs, March 2026).
The authority paradox: what really holds smaller brands back
The paradox fits in one sentence. For a language model to recommend a company, credible outside sources need to be talking about it already. But for anyone to talk about it, the company needs to be visible already.
NP Digital, Neil Patel's agency, asked 812 respondents to name the single biggest obstacle a smaller brand faces when trying to appear in AI-generated answers (NP Digital, October 2026).
Biggest obstacle | Share of respondents |
|---|---|
Limited brand recognition | 32% |
Few independent mentions or citations | 24% |
Insufficient original data | 13% |
Limited content resources | 12% |
Weak technical foundation | 8% |
Unclear measurement | 5% |
Difficulty choosing topics | 2% |
Other | 4% |
These are marketers' perceptions. Do the measurements agree? Yes, and clearly.
Muck Rack and Generative Pulse analyzed more than 40 million AI responses and 207 million citations in 2026. More than 84% of cited links come from earned media or third-party sources (journalism, government sites, research, encyclopedias, other companies' content). Brands' own sites account for 15.2% and paid content for 0.3% (Muck Rack, October 2026).
In a study of 75,000 brands, Ahrefs found that brand mentions across the web correlate at 0.664 with visibility in AI Overviews, compared with 0.218 for backlinks (Ahrefs, May 2025).
In other words, what others say about you weighs more than what you publish on your own site. Correlation isn't causation, but the pattern holds from one study to the next.
The opening: the more specific the question, the better a small business's odds
The same NP Digital survey asked respondents whether they felt able to produce a more useful answer than a larger competitor, by question type (NP Digital, October 2026).
Question type | Confident | Not confident | Net gap |
|---|---|---|---|
A customer's specific situation | 63% | 14% | +49 points |
Detailed implementation | 58% | 16% | +42 points |
Comparison with real trade-offs | 47% | 20% | +27 points |
Broad category question | 27% | 35% | -8 points |
Confidence drops as questions get broader and turns negative on category questions. That fits the authority paradox. For "best accounting software," AI has thousands of big-brand mentions to draw on. For "which accounting software for a Quebec construction subcontractor that bills with contract holdbacks," the pool of genuinely useful sources shrinks fast.
NP Digital also asked respondents to rate, out of 5, their firsthand expertise and how well larger competitors already cover several topic types (NP Digital, October 2026).
Topic type | Firsthand expertise (out of 5) | Coverage by larger competitors (out of 5) |
|---|---|---|
Niche use cases | 4.3 | 2.6 |
Implementation details | 4.2 | 3.1 |
Local or regional considerations | 4.1 | 2.7 |
General how-to questions | 4.1 | 4.3 |
Emerging trends | 3.1 | 3.4 |
Niche use cases and local considerations combine strong expertise with thin competition. General how-to content is already crowded. For a Quebec business, the local angle is concrete: provincial rules, seasons, distances, grant programs, market vocabulary, French-language content. Those are details large U.S. sites rarely cover.
The questions big brands answer poorly
NP Digital asked the 812 respondents to rate how well the largest brands in their category answer different types of customer questions (NP Digital, October 2026).
Question type | Big brands answer poorly |
|---|---|
Pricing, total cost and trade-offs | 50% |
Specialized use cases and product limitations | 46% |
Implementation steps, timelines and requirements | 43% |
Local or regional considerations | 38% |
Common customer mistakes and how to avoid them | 35% |
Specific industry needs | 29% |
On pricing, 31% rate big brands' answers somewhat poor, 19% very poor, and only 8% very good. One important nuance: specific industry needs are the one area where big brands hold up relatively well (32% rate them somewhat good). That's not where a small business has its best shot.
In practice, a page that shows realistic price ranges, explains what drives the cost and says frankly who the offer isn't for answers exactly what the big brands avoid.
You already have more evidence than you think
Authority is often associated with expensive studies. The data suggests a different starting point. Here are the types of evidence smaller brands say they could use publicly (NP Digital, October 2026).
Evidence available | Share of smaller brands |
|---|---|
Customer questions, objections and recurring feedback | 81% |
Practical lessons from delivering projects or services | 69% |
Customer examples or testimonials, with permission | 53% |
Interviews with internal experts or practitioners | 44% |
Original research conducted by the organization | 15% |
Aggregated, anonymized customer data | 11% |
Put that table next to another number: original research ranks first among the formats considered most effective in AI search (82%, see below), yet only 15% of smaller brands say they can produce it. The most realistic raw material is the questions your customers already ask your sales reps, your support team and your inbox.
Asked where to start, respondents rank improving existing specialist pages first: expected impact of 4.1 out of 5 for an effort of just 1.7 out of 5. Detailed comparison content has the highest expected impact (4.8) but also takes the most effort (3.2). An original research study scores 4.0 on impact for 4.1 on effort (NP Digital, October 2026).
What gets a source cited: perception versus measurement
Plenty of claims circulate about AI "ranking factors." It helps to separate what marketers believe from what studies measure.
First, perceptions. NP Digital asked 500 marketers and business owners to rate the importance of 13 factors for AI visibility (NP Digital, May 2026).
Factor (perceived) | Importance |
|---|---|
Brand mentions | 94% |
Reviews and sentiment | 91% |
Brand and entity authority | 87% |
Traditional SEO strength | 85% |
Structured data and schema markup | 83% |
Original research and data | 81% |
Trusted domain authority | 77% |
Content thoroughness | 70% |
Content freshness | 69% |
Author expertise | 68% |
Page speed | 15% |
Community engagement | 8% |
Backlinks | 3% |
Now set those perceptions against measurements:
Brand mentions. The perception (94%) matches Ahrefs' measurements (0.664 correlation) and Muck Rack's (more than 84% third-party citations).
Structured data. Here perception and reality part ways. In an Ahrefs test, adding schema markup produced no significant increase in citations in ChatGPT (+2.2%) or Google's AI Mode (+2.4%) (Ahrefs, May 2026). Google also says no special schema is required. Markup is still useful for rich results, but it isn't a shortcut to citations.
Backlinks. Respondents consider them almost irrelevant (3%). Measurements are more nuanced: their correlation with AI visibility (0.218) is weaker than mentions, but not zero, and links remain an SEO pillar that opens the door to AI Overviews.
Relevance. In an analysis of 82 factors, NP Digital ranks relevance first among the signals that influence ChatGPT recommendations (0.91), ahead of brand mentions (0.87), reviews (0.61) and authority (0.52) (NP Digital, December 2024).
What about the authority of the site that mentions you? NP Digital tracked 20 sites under two years old and published 300 articles recommending their products on sites of varying authority. ChatGPT's average likelihood of recommendation was 0.84 for DA 0 to 20, 0.79 for 21 to 40, 0.91 for 41 to 60, 0.85 for 61 to 80 and 0.87 for 81 to 100 (NP Digital, July 2025). The agency acknowledges the sample is small. Its practical takeaway: earn more relevant mentions rather than chasing only the highest-scoring sites.
Which content formats does AI cite most?
NP Digital surveyed 500 marketers and business owners about how 11 formats perform in AI search (NP Digital, May 2026).
Format | Reported performance |
|---|---|
Original research | 82% |
Comparison content | 76% |
Rankings and best lists | 57% |
FAQs | 41% |
How-to content | 39% |
Community and forum participation | 28% |
Generic blog posts | 25% |
Definition and explanation pages | 22% |
Opinion and thought leadership | 16% |
Product pages | 14% |
Video | 2% |
Two caveats. First, the source page doesn't define exactly what the percentage measures: these are self-reported results. Second, some figures contradict direct measurements. Video gets 2%, yet YouTube is the most cited domain in Google's AI Overviews (22.9% of mentions among the 50 most cited domains in September 2026, per Ahrefs), and YouTube mentions are the signal most correlated with AI visibility in another Ahrefs study (Ahrefs, December 2025).
Comparison content is even more interesting. It's rated highly effective (76%), but in a separate NP Digital survey of 1,000 marketers, comparison pages account for only 2.5% of reported citations, far behind listicles (21.87%) and articles (16.68%) (NP Digital, August 2026). Neil Patel reads this as more of a supply problem than a demand problem: few companies publish real comparisons. For a small business, that's an opening.
Does AI-referred traffic convert better?
Volumes are still modest, but visit quality looks higher. NP Digital reports the following averages, aggregated across 60 B2B and B2C campaigns (NP Digital).
Metric | Traditional traffic | AI-referred traffic | Difference |
|---|---|---|---|
Average conversion rate | 0.72% | 5.97% | 8.3x higher |
Average time to conversion | 8 days | 3 days | 62% faster |
Revenue per visitor | $2.56 | $18.04 | 7x higher |
Ahrefs sees the same pattern on its own site: AI search accounted for 0.5% of visits but 12.1% of signups, a conversion rate 23 times higher than organic search (Ahrefs, June 2025). These figures come from digital marketing companies; your results will depend on your industry. One plausible explanation: someone who clicks a source cited by AI has often already read a comparison and arrives ready to act, not to discover the topic.
What Google requires (and doesn't) for AI Overviews
Google Search Central's official documentation, updated in December 2025, is clear (Google):
there are no additional requirements or special optimizations to appear in AI Overviews or AI Mode;
to be shown as a supporting link, a page must be indexed and eligible to appear in Google Search with a snippet;
no special schema.org structured data is needed;
no new machine-readable files, AI text files or markup are required.
Ahrefs' data on llms.txt points the same way: across 137,210 domains analyzed, 97% of llms.txt files received no requests in May 2026 (Ahrefs, June 2026).
The technical work that matters is the classic kind: let crawlers explore the site, make sure important content is in the page's HTML, don't block snippets, and keep markup consistent with visible text.
Does AI-written content hurt GEO?
Google doesn't ban content produced with AI's help. Its position, published in February 2023, is to reward helpful content however it is produced, while treating automation used primarily to manipulate rankings as spam (Google Search Central). In an analysis of 600,000 pages, Ahrefs found no correlation between the share of AI-generated text and Google rankings (0.011) (Ahrefs, July 2025).
Watermarks do exist: in October 2024, Google DeepMind open-sourced SynthID Text, a tool that invisibly marks generated text (MIT Technology Review). But no public source shows a search engine demoting text because it carries a watermark.
The real risk lies elsewhere. Generic text that rephrases what already exists adds nothing a language model couldn't produce on its own, so it has no reason to be cited. What makes the difference is what AI can't make up: your real prices, your client cases, the mistakes you've seen, your numbers.
How to structure a page so AI picks it up
A page with good odds of being cited answers a decision, not just a question. Here's the structure we recommend:
The direct answer, in the first paragraph, in one or two sentences.
The why: what justifies that answer.
The conditions: situations where the recommendation changes.
Options and trade-offs, with price ranges where possible.
A concrete example from a real project, including its limits.
The next question the reader will ask, with a link to the answer.
Add subheadings phrased as real questions, tables for comparisons, update dates and linked sources. These are easy to extract, for a busy reader and for a language model alike.
A 30-day GEO plan for small businesses
Week 1: pick your pivot. Choose a priority customer profile and one specific question big brands handle poorly: pricing, limitations, implementation or local realities. Mine your emails, sales calls and support requests. It's the most accessible source of evidence (81% of smaller brands have it).
Week 2: gather the evidence. Frequent objections, lessons from projects, anonymized numbers, testimonials obtained with permission, a short interview with an expert on your team.
Week 3: improve before you create. Start with the specialist page you already have on the topic. It's the best impact-to-effort move according to NP Digital's respondents. Restructure it with the direct answer up top, add the evidence, a comparison table and internal links. Check that it's indexed and eligible for snippets.
Week 4: get it out there and measure. Pitch your insights to associations, partners, and trade or local media. Publish them where your customers look for information (LinkedIn, YouTube, specialized communities). Set up a routine to request reviews after every project. Then record your starting point in AI answers (see the next section).
How to measure your visibility in AI answers
Most smaller brands aren't measuring anything yet. According to NP Digital, 52% don't measure AI visibility at all, and only 2% can tie that visibility to revenue (NP Digital, October 2026).
What the brand can measure, at minimum | Share of smaller brands |
|---|---|
Nothing yet | 52% |
Whether it appears in AI answers | 48% |
Citations or referenced pages | 31% |
Qualified leads from AI discovery | 10% |
Revenue from AI discovery | 2% |
A simple way to start:
list 15 to 20 questions your customers actually ask;
test them monthly in ChatGPT, Gemini, Perplexity, Copilot and Google Search, and note whether your brand is mentioned, cited and described accurately;
in Google Analytics 4, isolate visits from chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com;
add "How did you hear about us?" to your contact form, with a "ChatGPT or another AI assistant" option.
AI answers vary from one run to the next. Muck Rack found that when the same prompt was run twice on the same day, ChatGPT repeated only 41% of the domains it cited the first time, versus 94% for Perplexity. Track trends over several weeks, not a single result.
How JB Impact can help
At JB Impact, we help Quebec small businesses with SEO and GEO: an SEO audit that includes AI visibility, rewrites of specialist pages, a mentions strategy and monthly tracking. To go further, read our strategic plan for answer engine optimization (AEO) and our look at AI in Quebec by the numbers.
Want to know what ChatGPT and Google say about your business? Book a free 30-minute consultation, no strings attached.
Frequently asked questions about GEO
What is the difference between SEO and GEO?
SEO aims to rank a page in search results to earn clicks. GEO aims to get a brand mentioned and cited in AI-written answers. The two work together: Google requires a page to be indexed and snippet-eligible before it can be cited in AI Overviews.
Why can a lesser-known small business get cited by AI?
Because big brands cover specific questions poorly. In NP Digital's survey of 812 respondents, 50% say big brands answer pricing and trade-off questions poorly, and 63% feel confident they can answer a customer's specific situation better. A page that documents that case with verifiable facts becomes a useful source for AI.
How many mentions does it take for ChatGPT to recommend a brand?
No reliable public study sets a threshold. What we do know: brand mentions correlate with AI Overview visibility (0.664 according to Ahrefs), more than 84% of AI citations come from earned media or third-party sources (Muck Rack), and in an NP Digital test, mentions on modest-authority sites also increased recommendations. Consistent, relevant mentions matter more than a magic number.
Which content formats does AI cite most?
According to NP Digital's respondents, original research (82%) and comparison content (76%) perform best, ahead of rankings (57%) and FAQs (41%). Generic blog posts score just 25%. Direct measurements, however, point to the importance of YouTube, which AI Overviews cite heavily.
Does the domain authority of the site mentioning you decide everything?
Less than people think. In an NP Digital test of 20 young sites, ChatGPT's likelihood of recommendation was higher for mentions on DA 41 to 60 sites (0.91) than on DA 81 to 100 sites (0.87). The sample is small, but it suggests a site's relevance matters as much as its score.
Do customer reviews affect AI visibility?
Marketers strongly believe so: 91% of NP Digital's respondents rate reviews and sentiment as important for AI visibility, and an NP Digital analysis of ChatGPT gives reviews a weight of 0.61. A routine of asking for reviews after every project is still the simplest move for a small business.
Do you need an llms.txt file or schema markup to appear in AI Overviews?
No. Google says no AI file, new markup or special schema is required. Ahrefs measured that 97% of llms.txt files received no requests in May 2026, and that adding schema didn't significantly increase citations.
How should you structure a page so a language model picks it up?
Give the direct answer in the first paragraph, then explain why, when the answer changes, the options and their trade-offs, a real example and the next question to ask. Use question-style subheadings, tables and linked sources.
Is AI-written content penalized?
Not in itself. Google rewards helpful content however it's produced, and Ahrefs found no correlation between the share of AI-generated text and rankings (0.011 across 600,000 pages). Generic text with no new data or experience simply has little reason to be cited.
Does traffic from ChatGPT convert better than organic traffic?
Early data suggests it does. Across 60 campaigns analyzed by NP Digital, AI-referred traffic converts at 5.97% versus 0.72% for traditional traffic. At Ahrefs, AI search drove 0.5% of visits but 12.1% of signups.
Methodology and sources
Every figure in this article was checked against its original publication, including the studies' charts. Most NP Digital data comes from online surveys of marketers and business owners: it measures perceptions, not the actual behavior of AI engines. Where possible, we set it against direct measurements (Ahrefs, Muck Rack, Google). NP Digital's domain authority test relies on a small sample (20 sites, 300 articles), as the agency itself acknowledges. NP Digital's page on AI traffic conversion gives neither a date nor a detailed methodology. This data moves fast: we'll update this article as new studies are published.
Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024
NP Digital, What Makes GEO Hardest For Smaller Brands?, October 2026
NP Digital, Where Marketers Believe Smaller Brands Can Compete, October 2026
NP Digital, Where Smaller Brands Have A Defensible GEO Opportunity, October 2026
NP Digital, Which Customer Questions Do Larger Competitors Answer Poorly?, October 2026
NP Digital, The Firsthand Evidence Smaller Brands Have Available, October 2026
NP Digital, The GEO Opportunities Smaller Teams Would Prioritize, October 2026
NP Digital, How Far Can Smaller Brands Trace The Value Of GEO?, October 2026
NP Digital, Which AI Visibility Factors Matter Most?, May 2026
NP Digital, Content Types That Perform Best in AI Search, May 2026
NP Digital, Which On-Site Content Types Earn the Most AI Citations?, August 2026
NP Digital, What Influences ChatGPT Recommendations, December 2024
NP Digital, Do High Authority Brand Mentions Influence ChatGPT?, July 2025
NP Digital, AI-Referred Visitors Convert 8x More and 62% Faster
Muck Rack and Generative Pulse, What Is AI Reading?, October 2026
Google Search Central, AI features and your website, December 2025
Google Search Central, Google Search's guidance about AI-generated content, February 2023
Ahrefs, brand mentions and AI Overviews, May 2025
Ahrefs, AI brand visibility correlations, December 2025
Ahrefs, 38% of AI Overview citations come from the top 10, March 2026
Ahrefs, most cited domains in AI Overviews, September 2026
Ahrefs, schema markup and AI citations, May 2026
Ahrefs, llms.txt study, June 2026
Ahrefs, AI-generated content doesn't hurt rankings, July 2025
Ahrefs, AI search traffic conversions, June 2025
MIT Technology Review, Google DeepMind is making its AI text watermark open source, October 2024
About the author
Jessy Bédard is a full-stack marketer at JB Impact, a web marketing agency based in Quebec City. A Wix Legend Expert (among the top 1% of Wix experts worldwide, with more than 350 sites built) and a certified Google Ads Partner, he has helped Quebec small businesses with SEO, GEO and online advertising for more than 20 years.
Comments