How To Audit Whether AI Recommends Your Brand

Why One Snapshot Proves Almost Nothing Generation involves randomness, and retrieval can return different pages between runs. The same prompt asked twice in a row can produce different companies in different orders.

Where Third Party Coverage Fits Even after all of the above, most citations in a commercial category will point somewhere other than your site. That is not a failure of your optimisation, it is how the system weighs self interested sources.

One scheduling detail improves comparability more than it should. Run on roughly the same date each month rather than whenever somebody remembers. Retrieval behaviour and the freshness of competing sources both vary over a month, and a series taken at irregular intervals introduces variation that looks like a trend.

Month Two: Corrections and the First Rewrites The work should now be concentrated on the recurring sources from the baseline. Expect a list of listings claimed, details corrected and errors submitted, with names and dates attached.

Write Passages That Can Be Lifted Citation happens at passage level, not page level. A model attaches a source to a specific claim, which means the unit of work is a self contained paragraph that remains true and useful when removed from its surroundings.

This frequently produces the first result of the engagement, because access failures are total and fixing them can change answers within days. It should also be short. A fifty page technical audit at this stage is usually padding drawn from a generic template.

Decide What the Result Means Four outcomes, each pointing somewhere different. Absent everywhere with a clean robots file and no third party listings usually means an identity and coverage problem. Absent with a blocked crawler or an empty non-JavaScript page means a mechanical problem, which is the good news outcome because it is cheap.

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.

Present but described wrongly means a source problem, and the source list tells you which page to correct. Present and accurate on definitional prompts but absent on the who should I hire prompts means your category presence is fine and your commercial positioning is not corroborated anywhere independent.

Keep a dated note of what you observed each quarter, including behaviour that later turned out to be temporary. The value is not in the individual observations, most of which expire, but in noticing how fast they expire. A team that has watched three of its confident conclusions become wrong within a year develops the right amount of scepticism about the fourth.

Those recurring domains are the pages your category's answers are being built from. Visit each one, look for yourself, and note whether you are absent, listed with stale details, or llm seo filed under the wrong category. That list is your task list, and you did not have to guess at it.

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.

Weeks Three and Four: The Access Findings A technical report covering crawler permissions, what the relevant agents actually receive from your server, whether bot management is interfering, and what survives on your key pages with JavaScript disabled.

Bring one other person from the business, ideally from sales. They will spot inaccuracies in how you are described that a marketing reader skims past, and they will tell you within minutes whether the prompts sound like real customers. That second opinion costs half an hour and prevents the most common flaw in a self run audit, which is a set of questions written in the company's own language.

Most engagements are judged too late, on a final outcome that arrives after the point where anything could have been corrected. The first quarter has its own deliverables, and knowing what they are lets you tell early whether you have hired the right people.

This means a single answer is a sample. Being absent once is not evidence of a problem and being named once is not evidence of success, and treating either as a result is the most common analytical error in this field.

The third question matters most. A good answer names a cause, attaches a number and admits an alternative explanation. A weak answer describes activity in the language of effort without connecting it to anything observable.

You are unlikely to read all of it, and its presence changes the incentives entirely. An agency that knows the raw evidence ships with the report writes a different summary than one that knows it will not be checked.

One practical consequence of the variation between systems is worth planning for. If your customers are split across two assistants that behave differently, resist building separate programmes for each. The shared requirements account for most of the achievable outcome, and the effort spent on system specific tactics is usually better spent widening the number of third party sources that describe you correctly.