Skip to content
NessFlow
Menu

Find out whether answer engines can cite you, and whether they do

Google now answers above its own results, and ChatGPT, Gemini and Perplexity cite sources of their own. We read your citations in AI Overviews and AI Mode every day, across the queries you already track, and we test crawler by crawler what your server really serves them.

Access opens in waves: we email you when yours is ready.

Our numbers, with their provenance

  • 19
    AI crawlers audited measured
  • 11
    operators covered measured

Traffic no longer tells the whole story

A generated answer does not necessarily send a click: it cites, it summarises, it decides, and the reader sometimes opens nothing. The three situations below almost always precede the question “why do our competitors show up and we don’t?”.

  1. 01

    Your pages are closed and nobody decided that

    A rules file written for Googlebot five years ago now blocks agents that did not exist back then. Nobody chose it, and nobody re-read it either.

  2. 02

    You are cited, or no longer cited, and a salesperson tells you

    Without a regular measurement, your only signal is a prospect reporting that they “asked the AI” and did not hear your name. That is late, and impossible to verify.

  3. 03

    Your rank tracking measures a page that changed shape

    A rank is still a rank, but the answer now sits above it, citing sources of its own. First place under an overview that names three other domains is worth less than the first place you were buying last year.

What we measure

Three separate measurements, never merged

Access, declaration and citation are three different questions. Merging them produces a single number nobody can challenge, which is the same as a number nobody can use. Each carries its own date and its own denominator.

  • Access, judged on what is served (every agent published by the major operators is checked against your rules, then a probe verifies what your server actually returns. A crawler allowed on paper and turned away at the door shows up as exactly that, with its refusal count beside it.)
  • What your files declare (your robots.txt and llms.txt are read, then compared with what you believe you wrote in them. A directive no known agent understands is flagged as exactly that.)
  • Share of citations, surface by surface (AI Overviews and AI Mode are read every day across the queries you track, and real questions are put to the assistants. Each surface carries its own denominator: one query can cite five sources in an overview and none at all in AI Mode.)
  • A versioned readiness score (computed on explicit axes, with its scoring version travelling beside it. Two measurements separated by a change of rules are therefore never silently compared.)

Frequently asked questions

How do I know whether my site is cited in Google’s AI Overviews?

On every query you track, the daily reading opens the AI Overview served that day and records the domains it cites. The screen prints two numbers side by side: how many times the overview appeared, and how many times you were in it. That first number is the one missing everywhere else, and it decides everything: Google chooses each day whether to show the surface at all, so a fall in citations can be nothing more than a fall in overviews. Without the denominator you would read a collapse where the engine simply changed its mind.

What does llms.txt actually do?

It is a declaration, not a barrier: it points agents to your reference content. It binds nobody, and no operator has committed to reading it. We check that it exists, that it is valid and that it agrees with your robots.txt, and we say what it does not do, because mistaking it for a block is the most common error on the subject.

Does blocking an AI crawler remove me from answers?

Not mechanically, and that is the trap. An engine can cite a domain whose pages it has never crawled, based on what other sites say about it. Blocking reduces what it knows about you; it does not erase you. The audit therefore separates access (measurable on your side) from citation, measurable in the answers. They are two different decisions.

Does citation tracking cost another subscription?

No. AI Overviews are read from the results page your rank tracking already collects each morning, and AI Mode runs weekly on the same entitlement, so both arrive with rank tracking, across the queries you track. Questions put directly to the assistants are a separate module, whose cost is announced before the run starts. The access check, meanwhile, runs again on every scheduled crawl, and a missed run reads “not measured”, never zero.

The exact scope of this measurement

This page sells a measurement. Influencing a model is a different thing, and the right-hand column is not a technical limitation we plan to lift later: these are things no honest tool can promise.

What we cover

  • Agents published by the major operators, with their purpose and their declared handling of your rules
  • Your rules files, re-read and checked against the real reachability of your pages
  • Your citations in AI Overviews and AI Mode, read across the queries you track

Stated limits

  • No guarantee of being cited: we measure an outcome, we do not influence a model
  • A day with no AI Overview counts as “not measured” and stays distinct from “not cited”
  • No rank and no ranking: a generated answer produces neither, and inventing one would be fiction
  • No retroactive history: measurement starts the day you start it

Start by knowing what is reachable

An audit of your domain says within minutes which agents reach your pages and what your server serves them. The AI Overview reading follows with your rank tracking, across the queries you already track.

Access opens in waves: we email you when yours is ready.