Search "how to get cited by AI" and you'll find a small industry has sprung up around the answer: agencies selling "GEO audits," tools promising to track your "AI citation share," blog after blog with a numbered list of tactics. Almost none of them show you data. They show you a checklist.
That's a problem, because a real dataset now exists — and it doesn't fully agree with the checklists.
The 8,000-citation study
In August 2026, Search Engine Land published an analysis of 8,000 citations pulled from 57 queries, run repeatedly across four AI engines — ChatGPT, Google Gemini, Perplexity, and Google AI Overviews — using data from the Rankscale.ai platform. It's the closest thing the industry has to a controlled look at what actually gets cited, rather than what marketers assume gets cited.
Three findings stand out for a business owner deciding where to spend time and budget:
- Self-promotion is close to invisible. ChatGPT cited vendor blogs — a company writing about its own product or service — in under 1% of citations. Google's AI Overviews and Perplexity were a little more generous, at around 7%, but even there, brand-written content was a minor player next to independent sources. As the analysis put it plainly: overly salesy content likely won't rank or be cited.
- Third-party authority does the heavy lifting. Wikipedia accounted for roughly 27% of ChatGPT's citations. Reuters showed up in about 6%. Perplexity leaned on expert review sites like NerdWallet and Consumer Reports for close to 9% of its citations. Across all four engines, blogs — mostly independent, third-party blogs, not brand blogs — made up 38–46% of citations.
- Quality beats rank. The study found cases where a lower-ranking page with genuinely authoritative content got cited over a higher-ranking page that was less credible. Position in Google's organic results is a signal these systems weigh, not the only one.
This lines up with separate research from AI-search analytics firm AirOps, which analyzed 21,311 brand mentions across ChatGPT, Claude, and Perplexity and found that 85% of brand mentions in AI search come from third-party pages, not the brand's own website — and for top-of-funnel, awareness-stage queries specifically, mentions were 6.5 times more likely to originate from someone else's content than the brand's own.
Put the two studies together and the pattern is hard to miss: AI engines are, in effect, running their own version of "don't take a company's word for its own excellence." They're looking for someone else vouching for you.
What this means if you're a local or service business
If your business serves customers directly — a clinic, a trades company, a restaurant, a local retailer — there's a second layer worth knowing about, and it just became a lot more concrete.
In July 2026, Yelp confirmed a data-licensing deal with OpenAI, first disclosed in its February 2026 earnings materials. ChatGPT can now surface Yelp content directly in local recommendation answers — star ratings, review snippets, and links back to your Yelp listing — when someone asks it for a plumber, a dentist, or a place to eat. The deal is non-exclusive, meaning the same underlying incentive (your reviews feeding an AI answer) likely extends to other platforms over time.
Moz's research on local reviews and LLMs, published the same week, found that large language models don't just average your star rating — they read the sentiment and specificity inside individual reviews to build a picture of trustworthiness. A detailed four-star review ("the hygienist was gentle and walked me through everything") can carry more weight with these systems than a flat five-star review that just says "Good." Google's own 2025 Maps Trust and Safety report, published in April 2026, disclosed that the company removed 292 million policy-violating reviews and over 13 million fake Business Profiles in a single year — a reminder that manufactured review volume is actively being cleaned out of the exact data AI systems are starting to read from.
For a local business, that reframes "getting cited by AI" as less of an SEO project and more of a reputation-management one: respond to reviews, encourage detailed (not just five-star) feedback, and keep your Google Business Profile and other listings accurate — because that data increasingly feeds the answer, not just the map pin.
What doesn't move the needle
Based on the data above, a few common "GEO" tactics are worth being skeptical of before you pay for them:
- A dedicated "AI citations" or "as seen by AI" page on your site. It's self-published, first-party content — the exact category the 8,000-citation study found AI engines cite least.
- AI-citation "tracking" tools as a starting expense. Knowing your citation count doesn't change it. Budget for earning citations first; measure once you have a baseline worth tracking.
- Generic FAQ-stuffing aimed at AI Overviews. It can help with traditional search snippets, but the data shows engines increasingly reward substantive, quality content over keyword-matched phrasing.
- Buying reviews or review volume. The exact pattern Google's Trust and Safety enforcement is built to catch — and the sentiment analysis in Moz's research means fake reviews often read as fake to the model too.
What actually works, in order of evidence strength
- Earn coverage on independent, authoritative sites. Digital PR, expert interviews, and inclusion in industry roundups feed the third-party sources AI engines already favor. This is slower and less controllable than writing your own blog post, which is exactly why it carries more weight.
- Publish something genuinely original. Proprietary data, a real client case study with numbers, an original survey — the kind of content other sites want to reference. Citations from other publishers only happen if you give them something worth citing.
- Make your site technically legible to machines. Clear headings, logical page structure, and content that directly answers a specific question in plain language, in that order — before worrying about schema markup or an llms.txt file.
- Treat every review as public data, because it now is. Reply to it, encourage detail, and don't chase volume over authenticity.
None of this is a quick fix, and that's worth saying plainly rather than selling around: nobody has found a "hack" for earning third-party trust, because if one existed, AI engines built specifically to detect promotional patterns would learn to discount it — as ChatGPT already appears to be doing with vendor content.
Try this yourself
Before spending on any AI-visibility tool or agency, run this free check: ask ChatGPT, Perplexity, or Google's AI Overview a question a real customer would ask about your industry — not your brand name, the actual problem you solve. Look at what it cites. If your own site shows up, note where. If it doesn't, note who does — a competitor, a review site, a publication — and ask whether that's a relationship worth building before it's a subscription worth buying.
FAQ
Does submitting my site to an "AI directory" help me get cited?
There's no evidence for this in the available research, and it resembles the low-value directory submissions that stopped moving traditional SEO rankings years ago.
Should I write content specifically "for AI" instead of for readers?
No — the data shows AI engines cite substantive, independently-verifiable content, which is also what serves human readers. Optimizing for one increasingly means optimizing for the other.
Do I need an llms.txt file to get cited?
No. Google has stated it doesn't use llms.txt for this purpose, and none of the citation research above ties citations to its presence.
How long does it take to start showing up in AI answers?
The research doesn't give a fixed timeline, but because most citations trace to third-party coverage and reviews rather than your own new content, expect it to move on the same timescale as earning press coverage or a reputation — months, not weeks.
Is this different for B2B versus local businesses?
Yes. The 8,000-citation study found B2B queries drew more heavily on company blogs and sites (about 17% of citations) than consumer queries, while local and B2C queries leaned on reviews, YouTube, and Wikipedia. The reputation-and-reviews strategy in this article matters more for local businesses; the digital-PR and original-research strategy matters more for B2B.
Takeaway
The data says the same thing twice, from two different research teams: AI engines cite third parties, not your own marketing. If a "get cited by AI" pitch doesn't involve earning coverage somewhere you don't control, it's selling you the 15% that moves the least.
Not sure where your business actually stands with AI search — or traditional search, for that matter? Talk to Digital Roof about an SEO strategy built on evidence, not guesswork. If the honest answer is that your content or site structure needs work before AI citations are even worth chasing, we'll tell you that too.
