AEO, Answered: Real Results, Why Some Sites Get Cited, and What Tools Should Do

Is anyone seeing actual results from AEO?

Yes. On a property built from scratch, AEO took it from zero to about 1,800 AI citations in roughly 120 days. But the useful part is not the number, it is which moves produced it and which did nothing.

Three levers moved the needle, in order of impact. First, get into the index each engine actually reads: ChatGPT retrieves from Bing, not Google, so importing into Bing Webmaster Tools and turning on IndexNow was the single fastest unlock, new pages were picked up in days. Second, ship content as server-rendered HTML: the major AI crawlers do not execute JavaScript,¹ so a client-side-rendered page reads as blank to them. Third, write liftable, answer-first passages, so a model can quote one self-contained paragraph without the rest of the page.

What did not move it: llms.txt (worth adding as cheap hygiene, but I have no evidence it changed a single citation) and chasing Google AI Overviews, which is the slowest surface to budge. So if you are "doing AEO" and seeing nothing, the cause is almost always one of two things: you are not in Bing, or your content renders client-side and the crawler sees an empty page. Test it in ten seconds, disable JavaScript, reload, and read what is left. That is what the AI sees.

Why do some websites get cited consistently by AI while others don't?

Because citation is a different contest than ranking, and the consistently-cited sites win four specific things, none of which is "authority" in the Google sense. I have watched pages that rank well in Google get ignored by every AI engine, and thin pages get quoted constantly. The pattern, in order:

1. They are in the right index. ChatGPT cites from Bing; Claude's citations overlap Brave's top results about 87% of the time.² Most "invisible" sites are simply not in those indexes, so nothing downstream matters.
2. Their passages are liftable. The cited paragraph stands on its own, answer-first, roughly 120 to 180 words, quotable without the surrounding page. About 44% of citations come from the first third of the page,³ so a buried answer rarely gets pulled.
3. They have information gain. They state something only they can: first-party data, an original number, named experience. A model cites the cheapest correct source, and original beats paraphrase.
4. They are corroborated off-site. Brand mentions correlate with AI visibility about three times more strongly than backlinks,⁴ and brands are roughly 6.5 times more likely to be cited through third-party sources than through their own domain.⁵

Sites that "rank but never get cited" almost always fail #1 or #2. Fix the index coverage and the passage shape first; they are the cheapest wins and they gate everything else.

What should an AI-visibility tool actually do?

Most of them measure the wrong thing. They track your "share of voice" in AI answers, a vanity number that reports the symptom and never the cause. Knowing you appear in 6% of ChatGPT answers for your topic does not tell you why, or what to change next.

A tool worth paying for would diagnose, in this order:

1. Index and crawl truth. Are you actually in Bing and Brave, and can the bots read your HTML without running JavaScript? This disqualifies most sites and almost no tool checks it.
2. AI-crawler log monitoring. Surface real OAI-SearchBot, Claude-SearchBot, and PerplexityBot hits from your server logs. Crawler visits are the leading indicator; citations follow. It is the signal I watch first.
3. Passage-level citability. Flag the paragraphs that are not self-contained or answer-first, the ones a model cannot lift.
4. Off-site mention tracking. Where you are mentioned across third-party sources, because that dominates citation more than anything on your own domain.
5. Per-engine citation tracking with the actual cited URL, not just "your brand was mentioned."

A good tool tells you the cause and the fix. I run my own server-log monitoring rather than buy one, because most tools stop at the vanity dashboard. If a tool only shows share-of-voice, it is a scoreboard, not a mechanic.

Do software review platforms like G2 and Capterra influence whether AI recommends you?

Yes, and more than most vendors expect, because review platforms are third-party corroboration, and corroboration off your own domain is what actually drives AI citation. When someone asks an assistant for the "best [category] software," the answer is almost never lifted from a vendor's homepage. It is assembled from the roundups, directories, and review sites the engine already trusts.

The data here is lopsided toward off-site. Brand mentions correlate with AI visibility about three times more strongly than backlinks,⁴ and brands are roughly 6.5 times more likely to be cited through third-party sources than through their own domain.⁵ A review platform is one of the cleanest third-party signals you can earn: it is structured, category-specific, and already sitting in the index the assistant reads.

Two practical rules. First, presence across several platforms beats a strong profile on one, because multi-source corroboration is what survives when an engine cross-checks a claim. Second, you cannot fake it, the reviews have to be genuine, so the play is earning real ones rather than gaming a single page. Test it yourself: ask ChatGPT or Perplexity for the best tools in your category and read which domains it cites. If G2, Capterra, or a Reddit thread shows up and you do not, that is your next quarter of work.

Does Google's AI Overview run on an LLM, and does my Google ranking feed it?

Yes on both counts, and that makes AI Overviews the one AI surface where your classic Google ranking still matters. AI Overviews are generated by Google's Gemini models, and unlike ChatGPT (which reads Bing) or Claude (which leans on Brave), they retrieve from Google's own index. So the SEO work that earns Google rankings is the same work that makes you eligible for an AI Overview. This is the single place where SEO and AEO genuinely converge.

But eligible is not cited. Ranking gets you into the candidate set; it does not get you lifted into the answer. Across AI engines, only about 12% of AI-cited URLs rank in Google's top 10 for the original prompt, and roughly 80% are not in the top 100 at all.⁶ Even on Google's own surface, the Overview pulls the passage that best answers the sub-question, not simply the top blue link. So treat a strong Google ranking as the entry ticket, then win the citation with a liftable, answer-first passage and information gain.

One caveat specific to this surface: AI Overviews skew toward older, established pages, so it is the slowest engine to budge. If you need fast pickup, ChatGPT and Perplexity can move within days; AI Overviews often take much longer to reflect the same work.

Has AI traffic changed how you measure SEO success?

Yes, because the click is no longer the whole scoreboard. An AI answer can fully satisfy a user without ever sending a visit, so sessions and rankings now undercount your real reach. Watch only Google Analytics and a page that is cited constantly by ChatGPT can look like a failure. I measure three things clicks never show:

1. AI-crawler hits in the server logs. Real visits from OAI-SearchBot, Claude-SearchBot, and PerplexityBot are the leading indicator: crawl precedes citation, so a rise in bot activity is the earliest sign the work is landing. It is the number I watch first, before any citation appears.
2. Citations per engine, with the cited URL. Not "share of voice," but which of my pages got pulled, for which prompt, on which engine. Track them separately, because only about 11% of domains are cited by both ChatGPT and Perplexity,⁷ so a win on one is not a win on all.
3. Seed-prompt coverage. A fixed list of the real questions I want to own, re-run across engines, recording who gets cited. That list doubles as the roadmap of what to create or earn next.

Rankings still matter as a diagnostic, but they lie about citations. The honest measure now is three questions: did a crawler read it, did an engine quote it, and for which prompt. If you cannot answer those, you are still scoring the old game.