Can you guarantee we get cited?

No. Nobody controls these engines. What is measured is whether you were a candidate, whether the page was read, and whether the sentence was lifted, and each of those has a lever you can pull and a re-check that says whether it worked.

Does a schema change cause a citation?

No claim is made either way. A change is recorded between two runs and read against the band; if the intervals overlap, the report says inside the band rather than crediting the change. That is what the experiment is for.

Which crawlers do you check?

The ones that decide whether an answer can read you: the training crawlers, the index crawlers and the on-demand fetchers, each named on its row with the robots.txt rule or the response that decided its verdict. The free check runs the same list.

How is the number produced?

Every rate carries a 95% interval from resampling the prompts, not the responses, and a judge whose agreement with hand-labelled answers is reported rather than assumed. The methodology page carries the working, limitations included.

For SEO teams

Your channel changed shape, and here is the technical fix list.

Find out why AI assistants skip your pages, which sites they cite instead, and a fix list that proves each fix landed.

robots.txt read per crawler, plus a live fetch · no account · or read a sample report

AI crawler accessacme.com · checked 6 Sep · no account

4of 6crawlers can read acme.com · robots.txt read per crawler · live user-agent fetch per crawler

Crawlerverdictdecided bylive fetch

PerplexityBot · robots.txt allows it · live fetch 403 from the edge robots.txt says yes and the CDN says no. The block is a firewall rule, and this is the one check that can detect it; the blocker on Actions closes when a fetch as PerplexityBot returns 200.

Read the working first

Show your working to the most sceptical person in the room.

Every finding comes with the robots.txt line, the response code or the answer that decided it. Nothing is a score you have to take on trust.

Per crawler

robots.txt read as each bot reads it, then a live fetch under its user agent.

Per absence

Five reasons you were not in an answer, and the lever each one has.

Per fix

Each fix is re-checked by the same test that found it.

01 · Why you were absent

Absence has five reasons, and each has a lever.

Every answer you were not in is filed under why: not a candidate, ranked and passed over, cited without the name, answered from memory, or couldn’t check. The lever is on the row.

Answers

OverviewMissedHistory

Acme · run #23 · absent in 242 of 576 answers

ReasonAnswersLever

96 answers · 11 prompts · most on “CRM with pipeline reporting, no admin” The engine searched, and no Acme page came back in what it read. The docs article on pipeline reporting is not on the retrieval path for this wording. lever · a page that answers the question in its own words, on the crawl · Actions › Queue

02 · The blockers

Know a fix landed without re-auditing it yourself.

Each fix is re-checked by the test that found it: the crawl, the access check or the accuracy check. You stop re-auditing pages by hand to learn what really shipped.

Actions

QueueBlockersSheet

Acme · run #23 · 5 open · re-checked 6 Sep

BlockerImpactCleared

Says the wrong thing2 · the accuracy check

Can be read1 · the crawl

Can be identified1 · the crawl

Answers nothing1 · the crawl

closes on · the crawl one engine cannot read the site at all · nothing published above this lands on it what has to be true · 0 of 3 · CDN rule allows PerplexityBot · fetch as PerplexityBot returns 200 · Perplexity cites a page, next run no correction · a site change closes it

03 · Whose page wins

Whose page wins the citation, and the sentence it wins with.

Every site the engines cite for your questions, how often, and what it says about you. Those are the pages to win.

Sources

AllNo recordWrong claimNo replyCompetitor namedNamed as fallback

Acme · run #23 · 546 citations · 12 domains

Answers grounded431of 576

Citation share11%of 431 grounded

Gaps3of 12

SourceWhoseLeverageFinding

listicle · earned · read 6 Sep · cited by 3 engines A list of 9 CRMs for small teams. Acme is not on it, and the list is the shared source in 23 answers. Pitchopens a pitch to the editor named on the page · nothing sends itself

04 · What engines retrieve

Find the searches engines run that nobody tracks.

The searches engines really ran while answering, and how often, beside the questions you track. One that comes up every time and nobody tracks is the cheapest question to add.

Prompts

TrackedSuggestedFrom engine research

Acme · run #23 · 32 tracked prompts · 6 engines

recommended9named, not picked8competitor named10not named3no answer stored2

Promptstageoutcomeengines

from engine research41 queries engines really ran while researching Acme

acme crm pricing 2026 asked every time Track

acme vs hubspot small team asked in 3 of 4 runs Track

crm pipeline reporting without admin asked every time tracked

learn · named on 3 of 6 engines · Attio and Pipedrive named in 14 of 18 The category prompt Acme should own and does not yet. The docs article is cited under 9 answers and the brand is not in the sentence that answers.

05 · Did it land

Prove the fix moved the answer.

Every change is measured before and after. No schema change guarantees a citation; the experiment shows whether it earned one.

Answers

OverviewMissedHistory

Acme · 6 runs · weekly · 6 engines

rundatesuitenamedvs previous

#23 6 Sep 32 prompts · 6 engines · 576 answers 58% +6 · cleared the band

#22 30 Aug 32 prompts · 6 engines · 576 answers 52% +1 · inside the band

#21 23 Aug 32 prompts · 6 engines · 576 answers 51% +2 · inside the band

#20 16 Aug 30 prompts · 6 engines · 540 answers 49% series break · 2 prompts added

experimentsa change recorded between two runs, read against the band

runswhat you changedrecordedchangeverdict

mention rate 44% → 47% · 95% ci 40–48 against 43–51 · n 576 both runs Within the band. The markup was found on the next crawl and the blocker closed; whether it moved the rate is not something this sample can say. Run measurementnext run 13 Sep · 32 prompts × 6 engines · priced before it runs

06 · It fits what you run

Fits the reporting you already run.

Nothing to install and nothing to ask engineering for. The same tables reach your reporting, your analytics and your assistant.

REST APIself-serve scoped keys · every table

CSV exportevery screen, with its basis columns

GA4AI traffic joined at the property

Slackthe digest, after every run

MCP server22 tools · ask in Claude or Cursor

Integrations and the API →

Fair questions

The ones you were already going to ask.

If it isn’t here, the methodology is public, limitations included →

A citation is not a ranking. There is one answer, it is written from a handful of pages, and the page that supplies the sentence gets the credit. The audit checks the same things you already check; the measurement reads what the engines actually retrieved and quoted.

Start with the one check that outranks everything.

Which AI crawlers can reach your site, per bot and per purpose, with the robots.txt rule or the response that decided it.

No account · a shareable result · or read a sample report