How it works
From your web address to an action plan
AI Says is a measurement product, so it is worth knowing exactly what it measures, what it infers, and where it stops.
- 01
You enter your website
One field. We normalise the address, confirm it is a public site, and retrieve the homepage.
- No account required to start
- No credit card
- Works with or without www
- 02
We identify your business
Your business name, category, location and services are read from your own structured data, metadata and headings — then shown to you with the evidence for each suggestion so you can correct anything wrong.
- Organization schema, Open Graph, title tag
- Address from structured data where present
- Services from navigation and section headings
- 03
We audit your answer-readiness
Twenty-four checks across crawlability, metadata, structure, schema, entity clarity, answerability and authority. Every check returns a plain-English result and the evidence behind it.
- Checks that cannot run return NOT_TESTED, never a guess
- Weighted roll-up becomes your AEO Readiness score
- Each failure maps to a specific recommendation
- 04
We build your question library
The questions buyers in your category actually put to AI, generated deterministically from your confirmed profile across five intent types — so the same business always produces the same library, and month-to-month comparison means something.
- Discovery, recommendation, comparison, commercial, informational
- Location-qualified variants where you name a market
- Priority-ordered so the highest-intent questions run first
- 05
We observe AI answers — where integrations are connected
Each question is put to every connected AI platform and the answer is parsed for whether you were mentioned, whether you were recommended, in what position, which sources were cited, how you were described and which competitors appeared.
- Platforms without a connected integration show "Integration required"
- Nothing is estimated to fill a gap in platform data
- Parsing is deterministic, so the same answer is always judged the same way
- 06
You get your score and your plan
Seven component scores, three plain-English answers, a separate data confidence rating, and a prioritised list of fixes with what, why, how and expected impact.
- Priority reflects both weight and gap size
- Every recommendation traces back to a specific finding
- The full report sequences the work into 90 days
The audit
Four possible results, and one of them matters more than you'd think
Most audit tools score anything they could not test as a failure. That inflates the problem and buries the real ones.
The signal is present and correctly formed. Counts fully towards AEO Readiness.
Present but incomplete, or technically valid but weak in practice. Counts as half.
Missing or actively harmful — a site-wide noindex, a blanket crawler block, no structured data at all.
The check could not run: a timeout, a block, a page that does not exist. Excluded from the calculation entirely, with the reason shown.
Query intelligence
What are your customers asking AI?
Your question library is organised by intent, because a business can be highly visible for informational questions and completely absent from the ones that precede a purchase.
Discovery
Customers looking for who does this kind of work.
Recommendation
Customers asking AI who they should actually use.
Comparison
Customers weighing you against the alternatives.
Commercial
Customers close to buying — price, process, what to ask.
Informational
Customers checking who you are before making contact.
For each question, the full report shows whether you were mentioned, whether you were recommended, your position, whether your site was cited, and which competitors appeared alongside you.
See it in the sample reportHonest limits
What happens when a platform isn't connected
Rather than showing you a number we cannot support, AI Says shows you the gap.
Where no AI platform integration is configured, the platform table reads “Integration required” and no per-platform score appears. Five of the seven components are then estimated from measured website and entity signals, capped well below the maximum, and each one is tagged Estimated. The overall score is labelled Preliminary and data confidence drops to Low.
Competitive Share and AI Sentiment are not estimated at all — there is no honest way to infer them from a website — so they are excluded and their weight is redistributed.
Estimated component scores are derived from measured signals on your own website and public entity signals — not from live AI platform answers. Connect AI platform integrations to replace estimates with observed data.
Ready to see yours?
The free check runs the full website audit and builds your complete question library.
Run a free checkRead the scoring methodologyQuestions
How the check works
Your website address. Everything else — business name, industry, location, services — is read from your own site and presented for you to confirm or correct.
No. AI Says reads only what is already public: your homepage, robots.txt and sitemap.xml, exactly as any search or AI crawler would. There is nothing to install and no permissions to grant.
Because the question library is built from them. If your market is read as "New South Wales" when your customers are in Western Sydney, every question we test will be slightly wrong — and so will the result.
The affected checks return NOT_TESTED with the reason attached, and they are excluded from the AEO Readiness calculation rather than scored as failures. It is also a finding in itself: if we cannot read your site, neither can an AI system.
Confirming your details takes about a minute. The analysis itself usually completes in under a minute. Where AI platform integrations are connected, running the prompt library adds time proportional to the number of prompts and platforms.
Find out what AI says about your business
Enter your website and get your AI Says Score. No credit card required.