Two kinds of AI audit. And what to do when AI is putting customers off your brand.

Most of the brands we’ve worked with at Known & Cited so far have come to us with the same question. They want to know if they’re showing up in AI answers. When somebody asks ChatGPT, Google’s AI Overviews or Perplexity who to use, who’s good, who to consider, are they in the list? Or is everyone else in the list?

That’s the most important question for most brands. It’s also the question that drives most of the tools and dashboards now selling themselves as GEO platforms. We’ve built our share of that work, and it’s the right shape for most brands that come to us early. We call it a discovery audit because it’s about whether AI is finding the business at all.

But it’s not the only question worth asking. And this week we took on our first client where it wasn’t even the right one.

The shape of a discovery audit

A discovery audit assumes a buyer is somewhere near the top of the funnel. They want a recommendation. They don’t know all the players. They turn to AI to be told who to look at.

So we ask AI the questions that buyer would ask. “Best UK preconstruction consultancy.” “Top luxury concierge for HNW individuals.” “Who should I use for B2B sustainability copy.” Unbranded. Open. We measure who AI names, how often, in what order, and which sources AI is leaning on. We score the brand on a five-band scale we call AVS, from Ghost (not in the answer at all) up to Known and Cited (the answer’s go-to). And we hand the data over with a recommended programme of PR, content and earned media to move the number.

For most brands, that’s the right starting point. They don’t yet have a problem with what AI says about them. They’ve got a problem with whether AI mentions them at all.

The shape of a reputation audit

The brief we took on this week was the opposite. The business gets named in AI answers all day long. It’s already known. The problem is what AI says when it talks about it. There’s a reputation issue, much of it historical, some of it overcooked by a few persistent media pieces, some of it played out in places the business doesn’t pay much attention to. The buyer they care about isn’t at the top of the funnel anymore. They’re already considering this business. They’re deciding whether to buy into it.

So we change the question. Every prompt we fire at AI contains the brand name. We’re no longer asking “who should I use,” we’re asking variants of “what’s it like to be a member,” “is the company in good shape,” “what do people say about its leadership,” “are there better alternatives.” Consideration and decision queries, not awareness.

The score scale changes too. The five bands of a reputation audit run from At Risk through Exposed, Mixed, Trusted, and Authoritative. Same idea, different framing. We’re measuring whether AI is reinforcing a positive or negative view of a known business, not whether it knows the business in the first place.

What a reputation audit shows that a discovery audit doesn’t

Two things stand out about a reputation audit, and both turned out to be the bit our client cared about most.

The first is the sources view. When AI is asked about a named brand, it has to fetch its answer from somewhere. Sometimes that’s the brand’s own website. More often it’s a review site, a forum thread, an old media piece that has aged into the training data, or a community post somebody wrote four years ago. A reputation audit maps that. It tells you not just what AI says about your business, but where AI is going to learn what to say. For PR people, this is the bit that earns the brief. You can’t argue with the answer without knowing where the answer came from.

The second is the sentiment shift under pressure. We ran the audit twice. Once with neutral questions. Once with the same questions reframed to start from a negative position. The difference between the two reads is the brand’s actual exposure. A business that holds steady from neutral to negative is in good shape. A business that drops a tier or two when a buyer comes at it sceptically is carrying a reputation risk that won’t show up in a single-tone scan.

Is AI putting customers off your brand?

Plenty of well-known brands carry reputation issues they’ve half-forgotten about. A short bad patch from five years ago. A leadership change that was painful at the time. A pricing complaint that hung around on Reddit. None of it would matter much if your buyer wasn’t asking an AI about you. But they are now. And AI doesn’t have the courtesy to forget. It pulls from whatever’s in front of it, and if what’s in front of it is the old story, the old story is what your buyer hears.

That’s the question worth sitting with. Is AI putting customers off your brand? Most well-known businesses haven’t asked, and a fair number of them would not like the answer.

A reputation audit is how a business finds out what AI is actually telling that buyer. It also tells you what to do about it. Not generic PR advice. Specific moves, briefed against specific sources, with specific outcomes to chase.

If your business is well known and you’re not sure what AI is saying about it, you probably should be.

Be Known. Be Cited.