AEO measurement · Vol. 1 · 2026

Can AI find your brand?

Measurement and optimization for the AI answer engines — with the confidence intervals other dashboards quietly hide.

╴ Free check · no signup · results in ~5 seconds
╴ The leaderboard · updated hourly

What's getting checked, and who's setting the example.

Most-scanned
Anyone, anywhere
  1. 01hackernews.comPartial
  2. 02httpbin.orgPartial
  3. 03reddit.comPartial

Click any row to see that domain's most recent report. Want to see your domain on this list? Run a check above.

§A · What's at stake

The internet is asking ChatGPT about you. You can't see the conversation.

In 2026, AI assistants are how people start searching — for products, articles, apps, services, experts, stories, code, news. Whatever you're trying to be found for, the engines diverge sharply in what they cite, the existing dashboards hide their measurement uncertainty behind a single number, and most "AI SEO agencies" rebranded from classic SEO in the last three months. We were not the first AEO tool. We are the first that publishes its methodology, names the conflicts other scanners hide, and gives you a shareable report a non-specialist can read without a glossary.

AnchorSource
Only 11% of domains are cited by both ChatGPT and Perplexity. One-size-fits-all AI optimization is broken.Profound platform analysis of 680M citations
Cited in Google AI Overviews → +120% organic clicks per impression, +41% paid clicks. The lift is real and measurable.Seer Interactive
Domain Authority correlates only r=0.18 with Google AI Overview selection. Classic SEO is not enough on its own.wellows AI Overviews analysis (15,847 results)
80% of URLs ChatGPT cites do not appear in Google's top 100 results for the same query.averi.ai · B2B citation benchmarks
If you're B2B specifically: 89% of buyers use AI for vendor research; the top quartile of SaaS brands earns 8.4× more AI citations than the bottom.Averi B2B Citation Benchmarks 2026
§B · What you get

A printable lab report, not a vanity score.

Every check returns a one-page diagnostic at a permanent URL. Static-scan today; live engine probes across ChatGPT, Claude, Gemini, and Perplexity arrive next. Five things we surface that other scanners don't:

  1. 01

    The ChatGPT-User foot-gun callout

    GPTBot is OpenAI's training crawler. OAI-SearchBot is the search index ChatGPT actually reads at inference. When you've blocked one expecting the other — or taken down both via a wildcard rule — we name the conflict by its actual mechanism, not a generic warning.

  2. 02

    All 14 AI crawler user-agents, categorized by purpose

    Training crawlers vs search-index crawlers vs user-initiated crawlers. Most scanners check 3–4 bots and lump them together. We cover OpenAI (3 bots), Anthropic (4), Perplexity (2), Google, Meta, Apple, ByteDance, Common Crawl — with vendor docs cited.

  3. 03

    Honest llms.txt framing — no overselling

    Anthropic Claude Desktop and Claude.ai respect llms.txt. Google has explicitly confirmed it does NOT (Gary Illyes, Search Central Live July 2025). OpenAI is unconfirmed. Most scanners tell you to ship llms.txt and imply it moves ChatGPT visibility. We tell the truth.

  4. 04

    schema.org FAQPage flagged as the highest-leverage gap

    The Relixir 2025 study measured 41% citation rate for pages with FAQPage schema vs 15% without — a 2.7× lift. Most scanners surface FAQPage as one checkbox among many. We surface it as the single most leveraged thing you can ship this week.

  5. 05

    Findings ordered by estimated impact, not severity-flatten-to-checklist

    Other scanners give you a 50-item to-do ordered by how easy it is to flag. We order by what would actually move citation share if you fixed it — calibrated against 2026 empirical work on retrieval mechanics.

╴ The share artifact

Every report lives at a permanent canaifind.com/r/{slug} URL with an Open Graph preview rendering the verdict block as a static image. Screenshot it. Drop it in your team Slack. Quote-tweet it. Send it to whoever's going to ask. The reports are how this tool finds the next person who needs it.

§C · Why we're different

The dashboards hide what their data costs. We don't.

Sielinski 2026 — the most rigorous published analysis of AI-search measurement to date — concludes that "single-run visibility metrics provide a misleadingly precise picture of domain performance in generative search." Every commercial AEO dashboard hides its confidence intervals because if they showed them, most of their lift claims would be statistically indistinguishable from noise. We show ours. Comparison across the things that actually matter when picking an AEO tool:

Can AI FindProfound / Otterly / PeecAthenaHQThe agency you're considering
Confidence intervals on every metric✓ Always visible✗ Hidden✗ Hidden✗ Not computed
Methodology cites peer-reviewed work✓ Sielinski 2026, Schulte 2026
Doer / measurer structural separation✓ Audit-only tier existsn/a — measurement only✗ Same tier, both roles✗ Same vendor measures own work
Off-site placement execution✓ Reddit, G2, Stack Overflow, YouTuben/a — measurement onlyRecommendations onlyVaries; typically content-only
The ChatGPT-User foot-gun callout✓ Flagged with the fall-through rule named✗ Not surfaced✗ Not surfaced
Honest llms.txt framing✓ Anthropic yes / Google no / OpenAI unconfirmedOversellingOversellingOverselling
§D · What we measure

Six metrics, every one with a confidence interval.

The metrics align with peer-reviewed work on AI-visibility measurement. Every cycle, every cell, reported with bootstrap CIs and excluding unverifiable URLs.

MetricWhat it countsTypical CIAnchor
Inclusion Rate (IR)% of runs where the model mentions your brand at all.±3–5ppPooled over intent cluster × model × region per §4.2
Citation Coverage (CC)% of runs that cite a verifiable URL on your domain. Hallucinated URLs excluded.±2–4ppHEAD/GET verified per §4.3
Share of Voice (SoV)Your IR vs named competitors' IR on the same prompt set.variesPer-engine and blended
Position scoreWhen mentioned, where in the answer (lead / body / footnote).±0.4Weighted, 0–5 scale
SentimentPer-mention classification (positive / neutral / negative / mixed).±5ppWith classifier-agreement CI
Citation diversityHow many distinct URLs on your domain are being cited.variesDeep links count more than homepage-only
§E · Methodology

Read the math before you buy the math.

Single-run visibility metrics provide a misleadingly precise picture of domain performance in generative search.

That's Sielinski, 2026. Our entire methodology is built around that finding: every metric sampled, pooled across intent clusters, reported with a bootstrap confidence interval. The hallucinated URLs flagged and excluded. The model snapshots detected and re-baselined. The competitor SoV computed per-engine because the engines diverge so sharply.

Read the full methodology

§F · Run yours

Type your domain. Drop the result in your team channel.

That's how this tool finds the next person who needs it. Free forever for one domain a day. No signup; no card; no Cloudflare dot-grid; no purple gradient. Just the answer.

╴ Anonymized into our industry source-domain corpus · privacy