Otterly logged 1M+ AI citations. AthenaHQ says the average brand shows up in 17% of buyer prompts. Neither number is yours.

Two of the better-known names in AI-visibility tracking spent real effort this year on something this category has mostly lacked: actual data, reported honestly, instead of another opinion piece about "the future of search." Otterly.AI analyzed over a million AI citations across ChatGPT, Perplexity, and Google AI Overviews. AthenaHQ published a State of AI Search 2026 report built from its own tracked-prompt corpus. Both are worth reading. Both prove, at a scale neither of us can produce alone, that the problem this blog exists to talk about is real. And neither one, by design, can answer the one question that actually matters to you: does AI recommend your brand, specifically, or someone else's?

What Otterly's report found

Otterly's "AI Citation Economy" report examined over 1 million AI citations gathered across ChatGPT, Perplexity, and Google AI Overviews between January and February 2026. Two findings stand out. First, community platforms — Reddit and Quora chief among them — captured 52.5% of citations, edging out brand-owned domains at 47.5%. When an AI engine backs up an answer with a link, it's now slightly more likely to point at a forum thread than at the company's own website. Second, and more actionable: 73% of sites the report examined carry some technical barrier blocking AI crawler access — a blocked robots.txt rule, a Cloudflare edge block, or similar. Nearly three out of four sites in that sample can't be read by the crawlers deciding whether they get cited at all, whatever their content actually says.

(Worth disclosing plainly: these figures come from WebSearch summaries of Otterly's own published report, not a fetch of the report itself — a direct attempt to load otterly.ai's report page 403'd from this sandbox tonight, the same standing block every competitor-domain fetch has hit all week. Multiple independently phrased searches converged on the same numbers, which is the corroboration bar this blog holds itself to when a primary fetch isn't available.)

What AthenaHQ's report found

AthenaHQ's State of AI Search 2026 report works from a different angle — not citation volume, but presence. Its headline finding: the average brand appears in just 17.24% of the buyer prompts it's tracked against, while the top-performing brands in the same categories reach 56.71%. That's not a small gap — it's visibility concentrating hard around a small group of brands, with most everyone else sitting well under one-in-five. The report also ties this to the traffic side of the story: AI-generated direct answers, which increasingly skip the click-through entirely, have contributed to search-traffic declines of 30-50% for many publishers since AI Overviews and similar features rolled out.

Same disclosure applies here — these numbers are WebSearch-consensus across summaries of AthenaHQ's own published report, not a direct fetch (also 403'd tonight), corroborated across independent sources rather than taken from one.

Why an aggregate report can't answer your question

Read both reports back to back and the honest limitation is structural, not a flaw in either vendor's work: a report built from a million-citation corpus or a whole tracked-prompt dataset is, by construction, an average. It tells you what's true across thousands of brands and prompts at once. It cannot tell you whether ChatGPT names your brand when someone asks for the best tool in your category, because that's not the question the report was built to answer — it was built to answer "what's true about the category," not "what's true about you." A brand sitting comfortably above AthenaHQ's 17.24% average could still be losing every single buyer question that actually matters to its business; a brand below it could be winning the three questions it cares about most and simply weak everywhere else. The average doesn't distinguish those two brands. Only a scan of your own domain does.

This is the same limit we've written about with our own curated study: an average score across 18 well-known SaaS brands came out to about 42/100, and that number is genuinely useful as a benchmark — but it was never meant to substitute for your own real number. Category data sets the baseline. It can't run the scan for you.

What we built instead of a second PDF

Rather than publish our own one-time report to compete on volume with Otterly's million citations — a number we don't have and won't invent — we built two pages that answer a narrower, more honest version of the same questions, and never stop updating. The State of AI Citations reads every real, non-mock scan ever run through our own pipeline and shows, per engine, how often it actually cites the brand's own domain versus just naming it — plus the real third-party source domains those engines pulled from, the same "who gets cited instead of you" question Otterly's report raises at category scale. The State of AI Search does the AthenaHQ-shaped version: mean visibility score, the share of tracked brands sitting below our invisibility threshold, and the biggest movers, computed live across every domain anyone has ever scanned through the app — curated or not, free or paid.

The honest mechanical difference from a published report: both pages are marked to render dynamically on every request, reading the live store instead of a build-time snapshot, and both exclude any run where no engine returned real data — a demo scan or an API outage never gets folded into the average as if it were an observed result. A category report is accurate on its publish date and ages from there. Ours is accurate right now, because it's recomputed every time someone loads it.

The number only you can get

Both reports are worth your time — Otterly's crawler-barrier finding pairs directly with our own field guide to the nine AI crawlers that decide whether you can be cited at all, and AthenaHQ's concentration finding is the exact dynamic behind why AI names a specific rival, not just a vague non-answer, when it skips you. But reading either one tells you about the category. It doesn't tell you about your domain.

Run the free scan and get the number neither report was built to give you: 25 real buyer questions, put to ChatGPT, Claude, Gemini, and Perplexity, read for whether you were named, ranked, and cited — or whether a specific competitor was. No signup, about a minute.