Your competitors show up in AI search results for one boring reason: the models have read about them more often, in more places, than they’ve read about you. There’s no hidden switch to flip. AI search visibility is mostly a mirror of how many brand mentions you’ve earned across third-party sources, review platforms, news coverage, YouTube, Reddit, comparison articles.
The models repeat what the web already says. If the web barely mentions you, AI barely mentions you either.
How AI answers actually decide which brands to name
An AI answer is assembled, not ranked. When someone asks ChatGPT or Gemini for vendors, the model combines what it memorized during training with what it pulls live at query time. Your name has to exist in at least one of those places (in the right company) to come out the other end.
Concretely, there are three places a model learns about brands:
- Training data. Everything the model absorbed from the public web before its cutoff. How often and how consistently your brand gets mentioned matters more here than how well any single page ranks.
- Live retrieval. When ChatGPT answers with search, it draws on third-party search providers and content from OpenAI’s publishing partners, then cites sources inline, that’s how OpenAI described it when launching ChatGPT search in October 2024.
- Query fan-out. Google’s documentation (updated December 2025) says AI Overviews and AI Mode can issue multiple related searches across subtopics to build one response, grounded in the regular Google index.
Google says standard search requirements still apply. Its docs state there are “no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.” No AI-only markup, no separate index to squeeze into.
So how does a model end up naming five vendors and not yours? Mostly through co-occurrence, which is how often your brand name appears next to a problem, a category, or a phrase like “best CRM for agencies” across everything the model has seen. Search Engine Land frames this as mutual information: repeated brand-topic pairings reduce the model’s uncertainty about what you actually do (July 2025).
Repeated mentions give the model more evidence to draw on. Recommending a vendor it has thin evidence for risks the model being wrong, so it names the brand that twenty credible sources already mention. Your competitor didn’t outsmart the model. It’s embedded in the model’s evidence base, and you’re not.
The seven reasons competitors beat you in AI answers
When I diagnose this for a brand, the gap almost always traces back to one of seven causes, and “we need better on-page SEO” is rarely the main one. Run these checks yourself; none takes more than an hour.
| Cause | How to check it | The fix |
|---|---|---|
| Thin third-party footprint | Search Google for “your brand” -site:yourdomain.com and count mentions on sites you don’t own | Earn mentions: PR, partnerships, review platforms, community participation |
| Inconsistent entity descriptions | Compare your About page, LinkedIn, Crunchbase, and Google Business Profile, do they even describe the same category? | Write one canonical brand description and use it everywhere |
| Missing from the sources models cite | Ask ChatGPT, Gemini, and Perplexity for “best [your category]” and note which sources get cited, are you on any? | Get onto those specific sources: review platforms, trade publications, comparison pages |
| Weak review footprint | Check volume, recency, and ratings on G2, Capterra, Trustpilot, or Gartner Peer Insights | Claim the profiles and build a steady review-generation habit |
| Content that never states specifics | Does your site contain numbers, prices, use cases, and comparisons, or adjectives? | Publish benchmarks, real pricing, and “who this is for” pages on your own site |
| Newer brand with little history | Are most of your mentions under six months old, with little coverage before that? | Stack compounding plays, reviews, communities, PR, and give it time |
| AI crawlers blocked | Check robots.txt for GPTBot, ClaudeBot, PerplexityBot, and Google-Extended | Make a deliberate call, blocking every AI crawler also blocks the models from learning about you |
Notice what’s missing from that list: keyword density, meta tags, schema gymnastics. The causes are almost entirely off-site. That matches what every major dataset on AI citation factors has found so far.
What the studies actually say about AI citation factors
The pattern isn’t theory anymore. The largest published dataset found brand web mentions correlate with AI visibility about three times more strongly than backlinks, and review presence alone moved median citation rates from 1% to 53.5%.
Start with Google. Ahrefs analyzed 75,000 brands in AI Overviews (May 2025) and found branded web mentions had the strongest correlation with being named, at 0.664. Backlinks? 0.218. Domain Rating? 0.326. In the same study, 26% of brands had zero mentions in AI Overviews, and top-quartile brands by web mentions averaged 169 AI mentions versus 14 for the next quartile down. The bottom half averaged zero to three, essentially invisible.
Ahrefs repeated the exercise across ChatGPT, AI Mode, and AI Overviews in December 2025. Brand web mentions again correlated at 0.66–0.71 across all three surfaces, and YouTube mentions were the single strongest factor overall at roughly 0.737. Domain Rating was mid-tier at best (just 0.266 for ChatGPT) and site content volume barely registered.
Then there’s the review angle. Seer Interactive’s May 2026 study of 804,491 AI responses across 1,926 brands found trust and review platforms are the number-two citation source once buyers reach the intent stage, 24.27% of citations, up from 1.51% at the awareness stage. Brands with no Trustpilot profile had a median citation rate of 1%. Brands with as few as 1–13 reviews hit 53.5%, and an active profile (median 81 reviews) added roughly another 25 percentage points.
Where do all these mentions live? Mostly in coverage you don’t own. Muck Rack’s May 2026 “What Is AI Reading?” report put earned and third-party sources at more than 84% of AI citations, with paid content at just 0.3%. Nearly 99% of everything AI engines cite is non-paid.
5WPR’s May 2026 analysis, built on Similarweb citation data from early 2026, found Wikipedia and Reddit together make up over 25% of ChatGPT citations in the US, while the WSJ, NYT, and Bloomberg didn’t crack the top 20. LinkedIn appeared in 14.3% of ChatGPT search responses. These are the third-party surfaces where brand lists actually get written.
And the academic baseline: the Princeton GEO paper (KDD 2024) tested nine content-level tactics across roughly 10,000 queries and found that adding citations, quotations, and statistics could lift visibility in generative engines by up to 40%.
A July 2026 critical survey reviewed 45 GEO studies and concluded that no reviewed technique produces a guaranteed outcome, and GEO rewrites can actually cut a page’s AI retrieval by up to 16%. Almost all the positive evidence, including the Ahrefs correlations, is correlational. Correlation isn’t causation. But when every credible dataset points the same direction, and the downside of acting on it is that you’ve built a more-mentioned brand, I’ll take that bet.
The catch-up plan I’d run
Start at the bottom of the cost curve. Entity fixes and review profiles cost almost nothing and compound quietly; PR and comparison placement cost more and take longer, but that’s where the durable advantage is.
-
Make your entity unambiguous. Entity consensus means every credible source describes your brand the same way, same category, same differentiators, same positioning. Audit your About page, LinkedIn, Crunchbase, directories, and Google Business Profile, write one canonical description, and roll it out everywhere. It’s a week of work that makes every other signal easier for models to read.
-
Claim and feed your review profiles. This is the highest-leverage cheap move in AI visibility right now, Seer’s data says going from no profile to a handful of reviews moves you from a 1% median citation rate to over half. Ask for reviews at moments of success, make leaving one take two minutes, and respond to all of them.
-
Get into the comparison set. AI answers lean hard on “best X” and “X alternatives” content. You can’t buy your way into all of it, but you can partner with adjacent vendors, contribute real data to industry roundups, and make sure analysts and comparison sites have accurate, current information about you.
-
Earn real coverage. Muck Rack’s numbers explain why digital PR keeps coming up in AI search: with 84% of citations coming from earned and third-party sources, press mentions are the supply chain of AI answers. Original data, surveys, and quotable expert commentary are what journalists cite, and what models repeat.
-
Show up where the conversations happen, carefully. Reddit threads, niche communities, and YouTube reviews all feed the models. But don’t treat any single platform as a moat: Reddit’s share of ChatGPT citations collapsed from roughly 60% to 10% in two weeks in September 2025. Participate genuinely, and spread out.
-
Publish specifics on your own site. The Princeton tactics apply to your pages too, and live retrieval still has to find something to cite. Real pricing, benchmarks, and clear “who this is for” pages give journalists, buyers, and AI answers something concrete to pass along.
-
Measure it like a brand metric, not a keyword. Keep a fixed set of 20–50 buying-intent prompts for your category and run them monthly across ChatGPT, Gemini, Perplexity, and AI Overviews. Track your share of mentions and which sources get cited. Expect movement in months, not weeks, this compounds like PR, not like a title tag.
Why building AI visibility takes time
There’s no shortcut here, and I’ve stopped pretending otherwise. If your brand isn’t in the model’s evidence, no amount of on-page optimization will make a model name you. The tools selling “GEO formatting” are optimizing the last mile of a journey you haven’t started.
That’s not just my grumpiness, the July 2026 survey of 45 studies found no technique reliably gets you retrieved in the first place. Formatting helps a page the engine already found. Becoming findable means becoming mentionable.
And mentionable is unglamorous. Reviews, PR, communities, partnerships, content with actual numbers in it, patience. That’s brand-building with a measurement layer bolted on, and calling it “GEO” doesn’t make it faster.
The flip side is that the moat is real. Here’s what I mean:
- A mention network can’t be bought in a week the way links could in 2010.
- Competitors who stack earned mentions now keep compounding while you deliberate.
- Every model update re-checks the same third-party web, so the advantage persists across model generations.
If your competitors show up in AI answers and you don’t, they didn’t find a trick. They did the boring work, or they’ve simply existed long enough for the work to accumulate. The good news is the work is knowable now, it’s checkable, and starting this quarter beats starting next year.
FAQ
Why don’t I show up in AI search results?
Because the models don’t have enough consistent, third-party evidence about you to risk naming you. Your own website doesn’t count for much, models lean on reviews, news coverage, communities, and comparison content. Thin mentions across those sources equals thin AI visibility.
What determines who appears in AI answers?
Mention frequency and consistency in the model’s evidence, live retrieval, co-occurrence with the topic, and the cited sources themselves. Review platforms and earned media dominate citations, while domain authority measures like backlinks correlate far more weakly.
How do competitors get cited in AI answers?
Same inputs, accumulated over time: press coverage, review volume, appearances in “best of” lists, community presence, and YouTube. Nothing exotic, they’re just embedded in more of the sources models read than you are.
Does digital PR help AI visibility?
Yes. It’s the most direct lever. With more than 84% of AI citations coming from earned and third-party sources (Muck Rack, May 2026), coverage that names your brand next to your category feeds the exact inputs models use.
How long does it take to show up in AI answers?
Timing is uncertain: answers grounded in live retrieval can pick you up within weeks of new mentions, while training-data-heavy answers lag longer. Treat three to six months of consistent mention-building as the realistic horizon.
Do backlinks still matter for AI answers?
For getting pages retrieved, yes. For getting brands named, they’re second-tier, Ahrefs found backlinks correlated at 0.218 with AI Overview visibility versus 0.664 for branded web mentions. Do both, but don’t confuse the two jobs.
Sources
- Ahrefs (An Analysis of AI Overview Brand Visibility Factors (75K Brands)) https://ahrefs.com/blog/ai-overview-brand-correlation
- Ahrefs (Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews) https://ahrefs.com/blog/ai-brand-visibility-correlations
- Seer Interactive (Study of 800K AI Responses: How Reviews Shape Brand Presence in AI Search) https://www.seerinteractive.com/insights/study-of-800k-ai-responses-how-reviews-shape-brand-presence-in-ai-search
- Muck Rack (What Is AI Reading? (May 2026 edition)) https://muckrack.com/blog/what-is-ai-reading-may-2026
- 5WPR (via PR Newswire) (Wikipedia and Reddit Now Drive Over 25% of ChatGPT Citations in the US) PR Newswire release
- arXiv (A Critical Survey of Generative Engine Optimization (2023–2026)) https://arxiv.org/abs/2607.14035
- Princeton / ACM KDD (GEO: Generative Engine Optimization) https://arxiv.org/abs/2311.09735
- Google Search Central (AI features and your website) https://developers.google.com/search/docs/appearance/ai-features
- OpenAI (Introducing ChatGPT search) https://openai.com/index/introducing-chatgpt-search
- Search Engine Land (In GEO, brand mentions do what links alone can’t) https://searchengineland.com/in-geo-brand-mentions-do-what-links-alone-cant-459367
