We ask the questions your customers ask.

We do not ask AI about you directly. We work out what your business is from your own website, then ask the questions a real customer would type when they are trying to choose a provider, like “Which data and analytics firms in Australia would you recommend?”

We generate several of these, covering different buyer intents: a shortlist ask, a specific-need ask, a comparison, a track-record ask. Each one is phrased so a helpful assistant would answer by naming real businesses.

Real assistants, browsing the live web.

We put those questions to real, web-grounded AI assistants, the ones your customers actually use, not a model reciting old training data. Today that is ChatGPT (with live web search) and Perplexity. Google's AI Overviews, Gemini and Claude are being added.

Every report names exactly which assistants tested you, and on what date. We never present “AI” as one universal system when it is not.

A measurement, not a single guess.

AI answers vary run to run. Asking a question once gives a number that jumps around, which reads as a guess. So we ask each question several times on each assistant, take the majority, and report a consistency rating alongside the result.

A competitor only makes your report if it comes up repeatedly, not on one lucky mention. That is the difference between a diagnostic and a coin toss.

Only real, like-for-like rivals.

When AI names other businesses, we check each one is real, operating in your market, and a genuine same-scale alternative, with a working website. We drop names we cannot confirm, and we set aside firms far larger than you rather than pass an enterprise giant off as your rival.

We would rather show you three competitors we can stand behind than five we cannot.

Only the signals that apply to you.

We read your website the way an AI crawler would, checking whether your identity, services, service area, credentials and structured content are clear and verifiable. We also read your structured data, the machine-readable section where a business declares its official profiles, so we can see the Google Business Profile, directory listings and social pages you link to even when they are not obvious on the page.

We only score the signals that apply to your kind of business. A national B2B firm is never marked down for missing a Google Business Profile or local reviews it has no reason to hold. Where a signal does not apply, we mark it “not applicable”, with the reason, and never count its absence against you. If you do have it, you always get credit for it.

If the word “signal” is doing a lot of work there, this is what we mean by it and why clarity beats volume.

Four states, and the fourth is not a grade.

Every line on your scorecard carries one of four states. Three of them are readings. The fourth is an admission.

Most tools report what they found and quietly treat everything else as missing. That hides the difference that matters to you. A signal we checked and did not find is a gap you can close. A signal we could not check is a limit of the test, and calling those the same thing would make the report easier to write and far less use to read.

So Not found means the check looked and it is not there. Not assessed means the check could not look, and it is never counted against you. Your result on screen and the plan we send you use these same four words, so nothing quietly changes meaning between them.

Where the Check honestly stops.

It is a point-in-time test. AI answers change as the web and the models change, so your result can move between runs. We show you the date and the consistency so you can read it fairly.

We do not test every AI tool, and we cannot guarantee any assistant will recommend you. What we can do is show the gaps clearly, and measure what changes when you close them.

Common questions

See how AI sees your business.

Run the free Check.