When nobody answers a question well, AI tools cite whoever comes closest. Here are three simple experiments where filling those gaps got our clients cited in AI answers, and how to try the same on your own site.
Tom LightIf you’re starting to look at AI search, one of the most useful things to ask is simple: which questions are your customers asking where you don’t appear at all?
Those gaps are often the quickest wins available. Publish a clear, expert answer to a question nobody else has answered well, and you can become the source surprisingly fast.
This is generative engine optimisation (GEO) at its simplest, and none of it is complicated. We’re not trying to game AI search. We’re running small, practical experiments to understand how it works, and anyone can do the same. Below are three examples from our client work.
In traditional SEO, a new page can take months to rank. AI search can move much faster, for two reasons.
AI tools search in real time. When someone asks a question, tools like Perplexity and ChatGPT often run a live search and pick sources from the results. A new page that answers the question well can be picked up almost as soon as it’s indexed.
There’s often no competition. For many specific questions, no brand has written a proper answer. AI tools fall back on forums and generic articles, or cite no source at all. The first brand to answer well has a clear run.
The catch is that you need to know which questions you’re missing. That’s where prompt tracking comes in: running a fixed set of customer questions through AI tools every week and recording who gets cited.
Insurance brand · two new guides · cited in most AI answers within days
The brand wanted to be seen as the expert in its core product. Our research showed that meant covering the full range of questions its audience asked while researching, not just the popular ones. Prompt tracking found two of those questions where the brand barely appeared.
We wrote a guide for each, answering the question directly and then adding the detail someone comparing their options would want.
Within days of launch, the share of AI answers citing the site rose from 16% to 70% on one question, and from 5% to 62% on the other. The guides also performed well in Google, earning over 43,000 impressions in their first three months and ranking around position two for both questions.
Travel brand · one new article · cited in four out of five answers within two weeks
This one came from customer experience research. We looked closely at the questions real customers were asking and found two groups asking the same questions: people new to this kind of holiday, deciding whether it would suit them, and people who had booked and were planning their time.
No travel brand appeared in AI answers to their questions, so we published one clear article answering them.
Before the article went live, none of the AI answers to these questions cited the brand. Two weeks after launch, 79% cited the new article, and the brand had moved ahead of most of its competitors.
Research organisation · refreshed existing articles · cited in around half the relevant AI Overviews
Not every gap needs new content. In this case, the organisation’s established articles were losing traffic. Some of that was natural content decay, but AI Overviews were making it worse by answering questions before people needed to click. (If that sounds familiar, see why is my organic traffic dropping?)
Our thinking was that if the AI Overview is the first thing people see, being cited in it builds awareness and makes people more likely to click when they want more detail. So we updated the articles, added clear answers near the top, and restructured them so the key points were easy to pick out.
This took months rather than days, but the articles went from being cited in almost none of their AI Overviews to around half.
All three experiments shared the same ingredients:
Audience research first. Every question came from understanding what people actually ask, not a keyword list.
A question nobody was answering well. That’s what made the wins quick.
A clear answer near the top of the page, followed by real detail and expertise.
Tracking the specific questions. Overall visibility scores barely moved. We only saw the results because we tracked the questions we targeted.
We’re still early in AI search, and nobody has all the answers yet. That’s why everyone should be experimenting. A small test on a couple of pages might only move a handful of answers, but it tells you what works, why it works and what to do next. Those small results are what shape the bigger strategic decisions further down the line.
Being used as a source doesn’t always mean your brand is mentioned in the answer.
For the insurance brand, the site was cited in around two thirds of answers, but the brand was named in only 6% to 7%. Many AI visibility tools only count brand mentions, so a real win can look like nothing has changed.
Our next experiment is getting brands named, not just cited. That means publishing information only you can provide, such as your own data, a lesson from our missing price information experiment. It also means being mentioned by other trusted sites, one reason digital PR matters more than ever in the age of AI search.
Talk to your sales and service teams, look at your search data, and think about what an expert in your field would need to cover.
Ask those questions in ChatGPT, Perplexity and Google, or use a prompt tracking tool to do it at scale. Note where you’re missing, and where no competitor appears either.
If you have a page on the topic, especially one losing traffic, update it first. If not, write a page that answers the question directly.
Look for whether you’re cited and whether you’re named, not just your overall score
Note what worked and why, and use it to shape the next experiment.
Start small. Two or three questions are enough to learn something useful.
AI search is still new, and the brands getting ahead are the ones testing, learning and building on what works. Content gap analysis is one of the easiest places to start.
If you’d like help finding the gaps in your own AI visibility, talk to us. For the bigger picture, read zero-click search in 2027, why you need to track more than one AI model and AI search maturity: where brands are now, and how to move up.
We’ll track the questions your customers ask, show you where you’re missing and help you run your own GEO experiments.
