AI answer-engine measurement

Answer engines decide what you are.
Measure it.

Two questions, one place: can AI answer engines actually read your site, and do they name you when your buyers ask? Start with a free 9-check readiness audit.

No signup, no card. Every check it runs is shown, pass or fail.

That example is our own site, scored 80/100 — we publish the number we actually got, including the check we fail. We have no customers yet, so there are no testimonials on this page.

Three instruments,
one footprint

Being readable and being cited are different problems with different fixes. These measure them separately.

Free · live now

Readiness audit

9checks
8passed here
1to fix

Nine checks on whether answer engines can reach, parse and attribute your pages — llms.txt, AI-crawler access in robots.txt, structured data, extractable text, canonical and sitemap integrity. Runs on any public URL in a few seconds.

Run an audit
From $19/mo

Mention tracking

Weekly measurement of whether ChatGPT, Perplexity and Google's AI surfaces name you for the prompts your buyers actually type — and who they name instead.

See a sample report
Free tool

Crawler analytics

Drop in a log file and see which AI crawlers hit you, how often, and which are impostors. Nothing leaves your browser.

Open the analyzer

5%of a typical page's HTML
is text a crawler can read

Most of what you ship is invisible to an answer engine

Across the sites we audited for this page, visible text averaged about 5% of total HTML — the rest is markup, inline script and styling. An answer engine retrieving your page keeps the small part and discards the rest.

That is why a site can pass every conventional SEO check and still never be quoted: the substance is there for a browser and thin for a retriever.

Text a retriever keepsEverything it throws away

Reference data, free and open

The numbers we needed ourselves and could not find in one honest place. No signup — these exist to be cited.

Dataset
LLM API pricingPer-token pricing across the major providers with a cost calculator, updated from each provider's own pricing page.Open →
Model deprecationsRetirement dates for models already scheduled to be withdrawn, so a shutdown is not a surprise.Open →
Provider statusOne board for nine AI providers' status pages instead of nine tabs.Open →
GPU cloud pricesHourly on-demand GPU pricing across cloud providers, with the quote basis stated.Open →

Output price, per million tokens

34 models across 3 providers, from each provider's own pricing page. Verified August 2026.

Cheapest and most expensive models by output price
Gemini 2.5 Flash-Lite
Google
$0.4
GPT-5.6 Luna
OpenAI
$1.2
GPT-5.4 nano
OpenAI
$1.25
— most expensive —
GPT-5.4 Pro
OpenAI
$180
GPT-5.5 Pro
OpenAI
$180
Claude Opus 4
Anthropic
$75

All 34 models + cost calculator →

Pricing

The audit is free permanently. Where a paid thing is not built yet, it says so.

Always free

$0

Readiness audit

  • All 9 checks, every result shown
  • Any public URL
  • No account
Launching — waitlist open

$19/mo

Mention tracking

Not built yet

$19

Written fix report

  • Prioritised, per-page fixes
  • Ships after launch — join the list rather than pay now

Questions

Q1

What is an AI answer footprint?

Two separate things: whether AI answer engines can technically access and parse your site, and whether they actually mention you when someone asks about your category. A site can be perfectly readable and still never be cited, and the fix for each problem is different.

Q2

Is the readiness audit really free?

Yes. The audit runs on any public URL at no cost and shows every check it ran, including the ones you pass. The paid tier is a written report with prioritised fixes.

Q3

Do you store the sites I audit?

We keep the URL and the resulting score so we can show you change over time. The crawler log analyzer never uploads anything — it runs entirely in your browser.

Q4

How is this different from a normal SEO audit?

A normal SEO audit optimises for a ranked list of blue links. These checks optimise for being retrieved, parsed and attributed by an answer engine, which fails for different reasons — blocked AI user-agents, unparseable client-rendered text, missing structured data.