How To Audit Whether AI Recommends Your Brand

Measure Position Change in the Prompt Set This is the closest thing to an output metric that you can genuinely audit, because you own the instrument. Run a fixed prompt set on a fixed schedule under fixed conditions, and track four things:

That matters most for the facts that establish identity, because those are the facts that let scattered mentions of you resolve into one record. It matters far less for content, where the model is going to read the prose anyway and is reasonably good at it.

Long sections on activity that produced nothing, described in the language of effort rather than outcome. And the most reliable indicator, a report you cannot disagree with, because it contains no specific claim to test.

How to Judge Progress at Each Stage Use different measures at different points rather than asking for mentions from month one. At the end of month one, ask whether the baseline exists and whether access problems were found. At month three, ask whether listings are corrected and whether your own pages appear in citation lists at all.

Keep a record of what you predicted as well as what you measured. Writing down at the start of a quarter what you expect to move, and then reading it back at the end, is the cheapest way to find out whether your model of this channel is any good. Most teams never do it, which is why the same confident explanations survive for years without ever being tested.

Two asking who to hire or buy from for the thing you sell. Two describing the problem your product solves without naming the category. Two comparing named competitors. Two asking about a specific situation your best customers are in. One asking directly who your company is. One asking whether your company is any good.

Crawler access restored on a date. Listings claimed and corrected, with a count. Factual errors fixed on third party sources, with a count. Pages published that answer prompts your baseline showed were being answered badly. Reviews responded to.

Read the Source List Before the Prose Where citations are shown, list every domain and count how often each appears. This is the single most useful output of the whole exercise, and most people skip it because the prose is more interesting.

Expect the timeline to be uneven. Crawler access can change what an assistant sees within days, because retrieval happens at answer time. Identity consistency takes longer, since scattered mentions have to be re-crawled before they join up. Third party coverage is slowest of all and is the part you control least directly, which is exactly why it is worth starting on it before you need the result.

One warning about testing. If you fix something and immediately re-run a prompt in the same session, the assistant may repeat its earlier answer from context rather than retrieving afresh. Start a new session, and run the prompt several times, before concluding that nothing changed. generative engine optimization

One to Three Months: Listings and Corrections Claiming a directory profile, correcting an address, fixing a miscategorisation and responding to reviews all take effect once the platform publishes the change and the page is re-crawled.

What Moves the Timeline Category coverage is the dominant factor. A category with two thin comparison articles can move in a quarter. A category where every comparison page has been fought over for a decade may take a year to enter.

This section sounds procedural and it is the foundation of everything after it. A prompt set quietly edited between runs makes every trend line in the document meaningless, and it is the easiest way to manufacture improvement without doing anything.

Use the Soft Signals Deliberately Two free signals carry more information than their informality suggests. Add a how did you hear about us question to your enquiry form and read the free text monthly rather than the categories.

A frequently quoted comparison showing assistant referrals converting several times better than search came from a vendor selling the service, across 312 business to business brands. A widely shared claim about explosive referral growth rested on nineteen analytics properties. Both are legitimate observations and neither supports the confident generalisation usually attached to them.

The truthful answer is that different parts of the work move at very different speeds, and knowing which is which lets you judge an engagement at the right moment instead of the convenient one. generative engine optimization

What a Defensible Business Case Looks Like It states what cannot be measured. It reports inputs completed, with counts. It reports prompt set movement as fractions with visible run counts, split by intent. It includes the soft signals as anecdote clearly labelled as anecdote. It attributes every external statistic.

Three to Nine Months: Earned Coverage The slowest and most valuable part. Getting into the comparison articles, trade publications and community discussions that assistants actually cite depends on other organisations deciding to write about you, which no amount of budget reliably accelerates.