TLDR: Run an AI visibility audit while you are still shaping the strategy for a new brand launch or repositioning and use it as a market intelligence tool. For a new launch, it shows who already owns the category and which sources shape AI’s recommendations. For a repositioning, it shows how AI sees you and how much of that picture fits the story you want to tell.
Most brands have never audited their AI visibility. The few that have tend to treat it as a scorecard: a quick look at how they fare, filed away once the curiosity is satisfied. That is a missed opportunity.
Run the audit while you are still shaping a launch or repositioning and it becomes a strategic tool. It shows you what the public record already says, who owns the territory you want to conquer and what it will take to change the story.
AI assistants build their answers from everything published about a brand: its own website, press coverage, reviews, booking platforms, forums and curated guides. Before you commit, ask AI the questions you want to be the answer to. What comes back will show you the state of the market you are about to enter, the evidence used to judge your place in it and the ground your new positioning will have to win. You are not asking ChatGPT to invent your positioning, but to examine the information environment your brand will have to compete in.
I recently ran an AI visibility audit for a luxury hotel in London that wanted to move away from its traditional luxury image towards something more modern and relaxed. Its new identity, language, target media and competitors were already agreed, and the AI audit was commissioned to inform the new website copy – a very clever move that ended up testing the whole strategy.
Here is how an AI visibility audit can shape your repositioning, step by step, using the findings from this case study.
1. Build questions around the space you want to conquer
Test with the questions your ideal customers would ask when they need what you are selling. If you want to own the design-led luxury space, test prompts like “Which design-led hotels in central London would you recommend?” Leave any brand names out.
Run these alongside questions that name your brand and keep the two sets separate. Here is why:
The branded prompts are the named questions customers ask when they already know your brand: “Is X worth booking for a weekend?” or “How does X compare to Y?” They show how well AI understands your brand once you are part of the conversation.
The non-branded prompts are discovery questions. They describe a need without naming a brand: “What are the best boutique hotels in Mayfair for a couple’s getaway?” Here, AI makes the choice, drawing on whatever it can find online from the sources it considers authoritative.
The latter matter most when testing your brand’s positioning. They reveal which brands AI associates with the space you want before anyone has mentioned you.
The gap between the two sets is usually dramatic. For example, the hotel I audited appeared in 93% of the answers to branded prompts, but only 6.5% of the answers to discovery prompts. And when AI did describe it, it used the hotel’s own key phrases in less than 1% of answers.
2. Find out if you are described the way you want to be described
Most dashboards report whether a brand appears. For positioning, what matters is how your brand is described.
List the words you want associated with the brand and the words you want to leave behind. Then see how often each appears in AI’s descriptions.
For the hotel, “grand” appeared in two-thirds of descriptions and “opulent” in one in five. The word at the heart of its new positioning appeared in 3%. AI had not misunderstood the brand. It understood the version the brand was trying to leave behind.
A standard sentiment score would have missed this. Almost every description was positive, so on a dashboard the hotel looked fine. It was simply the wrong kind of positive.
The question is not “Are we visible?” but “Are we visible for the reasons we want to be?”
3. Find who’s the sheriff in town
Study every brand AI recommends for your target questions. Then compare that list with the competitors in your strategy deck.
They rarely match. The hotel I audited had benchmarked itself against ten contemporary lifestyle brands. AI recommended traditional grand hotels such as The Savoy and Claridge’s almost as often as the chosen peer set.
Even more surprising, half of AI’s design-led recommendations were hotels not on the list at all.
Customers do not categorise brands the way businesses do. Your competitive set is increasingly defined by the question being asked, not by the names on your competitor list.
This step also tells you whether the space is worth pursuing. A positioning can sound distinctive in a boardroom and still be crowded in the market.
4. Find the sources that shape the answers
Check which sources AI cites when it talks about you and about the brands that win in your place.
Here is what to look out for:
1) How little of the record you control. In my audit, only 21% of the citations came from the hotel’s own website. So, a new homepage alone will not move the other 79%.
2) How long old stories survive. AI was still describing a restaurant the hotel had closed some time ago in 165 answers, sourced from older third-party coverage. Outdated evidence does not expire on its own. It must be corrected at the source.
5. Separate the facts gap from the framing gap
Before building your positioning plan, ask why AI does not yet connect you with the space you want. There are usually two possible answers, and they lead to very different plans.
1) A facts gap means the proof does not exist yet. A hotel that wants to own wellness but publishes no spa menu, facilities or practitioners gives AI nothing to find. No amount of messaging will fix that. The brand needs something real to point to first: a new product, service, partnership or experience.
2) A framing gap means the proof exists but is described in the old language. The facts are there. The story around them has not caught up.
My audit exposed a framing gap. AI already mentioned the hotel’s daily social hour and pet-friendly stays, both strong signals of warm, relaxed hospitality, and described it as “playful” in a third of answers.
It did, however, wrap those facts in words like “grand” and “opulent”.
So, the raw material for the new positioning was already public. What needed to change was how it was described – by the hotel and by the sources AI trusts.
Sometimes the audit points to a third conclusion: the space is already owned so firmly by others that it is not worth chasing. And that is a far cheaper lesson to learn before the campaign than after.
6. Build the evidence plan
Now the positioning becomes a plan: desired association → required evidence → authority sources → PR activity.
For the hotel I audited, that meant retiring “grand” and “opulent” from its own language, repeating the new vocabulary consistently across releases, landing pages and captions, and getting into the sources that shaped the category. One curated hotel-discovery platform was cited 789 times in answers a competitor won, and 176 times in answers featuring the hotel. That is an easily fixed placement gap, and it belongs in the PR brief.
This is also where communications teams earn a place earlier in the process. The evidence a positioning needs is exactly the ground PR knows best.
7. Run it in every language your customers use
For brands selling into international markets, English is not enough.
My research this summer asked AI the same questions about London luxury hotels in English and Arabic, for a Gulf traveller audience. The recommendations were 79% similar across the two languages. The sources behind them were not: excluding booking sites and social media, only 28% of the websites cited appeared in both.
A repositioning can take hold in English while the old story survives in Arabic, built from sources the English plan never touched.
8. Then keep the audit as your baseline
The same audit becomes the benchmark for the repositioning itself. Re-run it quarterly and track the numbers that matter: discovery rate, the share of old and new language and where the evidence comes from.
Before launching the new story, find out which story the evidence already tells. That is what your new positioning will have to overcome.
