Everyone’s asking the same thing right now: is AI search about to change SEO, or is that overblown?
Our Head of Marketing, Rosie, put this straight to Tim, our Head of SEO. She didn’t let him off with the easy version.
Is ranking dead, is this just digital PR wearing a new badge, or can any of it even be measured?
Here’s what he said.
You’ve got one sentence to kick us off. What’s your top tip?
Tim: By 2027, brands will be recommended for reasons rather than ranked for keywords. The biggest lever becomes AI SEO: brand marketing that makes sure AI genuinely understands what sets you apart..
Why does this matter, and who should be worried about it?
Tim: It matters because the unit of competition has changed.
For 20 years, SEO meant ranking a page for a keyword. In AI search you’re not fighting for a position. You’re trying to be the brand the model recommends when someone asks for advice.
Almost all the tooling, reporting and habits still assume keywords and rankings. So a lot of teams are getting very good at a game that’s quietly being replaced.
It’s aimed at everyone eventually, but it’s most urgent for challenger brands and specialists. These are businesses that can’t out-spend or out-fame the category leader, but do have a genuine point of difference.
Take a considered purchase: finance, B2B software, professional services. If people there ask AI for a recommendation, it’s already happening to you.
Hang on, if brand’s the new lever, does ranking even matter anymore?
Tim: It’s two layers, not either-or.
The model doesn’t start from a blank page. Ask it for the best betting brand and it opens with Paddy Power, Sky Bet and Bet365, then researches around that. Brand shapes the answer before anything’s even been searched for.
But when it does search, it behaves a lot like a human would. It runs around 11 searches for a single question, pulling roughly 28 links each time. Most of its attention goes on the top few results.
A source in the top spot gets cited about 69% of the time. That halves at the 12th-to-13th cliff, from 32% to 15%. Being in the top 12 is still the game.
Depth matters too. Betfair’s the most-found bookmaker in our data: 513 pages surfaced against Bet365’s 226. That’s because Betfair’s Exchange product gives it a bigger back-catalogue of published content.
When five or more of a brand’s pages were surfaced, it got cited in around 90% of questions. With just one page, that dropped to roughly 52%.
So, brand sets what the model assumes going in. Rankings get you into the sources it reads and cites.
SEO’s necessary. It’s just not sufficient on its own anymore.
Can you give us an actual example, not just theory?
Tim: Take a wealth manager whose edge is genuine independence.
Nobody searches for “best independent wealth manager”. But ask AI which firms won’t just push me into their own funds. Independence becomes the exact reason it surfaces them.
You get recommended for a reason, not a keyword.
Isn’t this just digital PR with a new name?
Tim: Partly, yeah, and I’d rather be honest about that than pretend it’s magic.
The tactics overlap: DPR, getting talked about, being cited. What’s changed is the target.
DPR used to be about awareness, and an effective form of building high quality links odd link. Now it’s feeding the thing that recommends you.
Our own research says DPR won’t win you the generic “best X” queries, though. We ran an AI visibility study across UK betting brands, tracking who ChatGPT, Gemini and Perplexity actually name and cite. Roughly 67% of the sources were affiliate or comparison sites.
Ask a generic question and it collapses straight to the big affiliates and US-famous brands. Digital PR won’t win you those, and saying so out loud is what earns you credibility with an SEO crowd.
Where DPR earns its keep is the narrow, ownable niche.
Ask about spread betting and the model names Spreadex first and cites spreadex.com. Ask for best-odds-guaranteed racing and it cites StableBet, the BOG specialist, rather than generic Oddschecker.
The job is engineering a sharp, UK-legitimate topical association so you become the cited authority in a narrow, ownable category.
Vague positioning gets you nowhere.
Okay. What should an SEO go and do on Monday morning?
Tim: A few concrete things.
First, map the real questions people ask, not keywords. Pull them from People Also Ask, Reddit, sales calls and the language customers use, and accept it’s an estimate.
Second, run positioning probes. Ask ChatGPT, Gemini, Perplexity and Claude the buying questions your customers ask. Record who gets named, who gets cited, and for what reason. That’s your baseline.
Third, find your ownable niche. That’s the sharp, specific association you can credibly win, not the generic “best X” you’ll lose to fame and affiliates.
Fourth, build that association off the site: run digital PR, get cited by the sources the models trust, and send consistent signals about what you’re known for.
Then re-probe over time and see if it’s forming.
This all sounds like it needs a huge budget. What about smaller businesses?
Tim: Honestly, it’s the opposite. It’s where smaller businesses have the best shot.
You’ll never out-fame the market leader on the broad stuff, so don’t try.
But a small, sharp, credible niche, the one thing you’re genuinely the best answer for, is winnable. The giants usually can’t be bothered with it.
Small budgets should go narrow, not broad. That’s the whole point.
Can we trust what these models say, given the risk of hallucinations?
Tim: You can’t fully, no, and I wouldn’t pretend otherwise.
Ask the same question twice and you’ll get different answers. That’s exactly why you probe repeatedly and look at the trend, not a single result.
You’re building a directional picture, the way brand tracking always has.
If you’re after penny-perfect certainty, this’ll frustrate you. If you can live with “are we showing up more often, for the right reasons”, it works.
So how do you measure whether any of this is working?
Tim: Not with rank tracking. You measure share of model instead.
Across a basket of buying-intent prompts, how often are you named and cited versus competitors, and is the trend rising?
Then you accept it’s uncertain and incomplete, the way brand tracking or a TV campaign always has been.
So, you triangulate: probe results, branded search and direct traffic, referral traffic from the AI tools, assisted conversions.
You’re after directional confidence, not exact numbers. Measure it like a brand marketer, not a last-click performance marketer.
Last one. If someone reading this does exactly one thing today, what should it be?
Tim: Open ChatGPT this afternoon and ask it the five questions your customers ask before they buy.
See who it recommends, and why. That gap, between who it names and who you want it to name, is your SEO strategy for 2027.
Keep your rankings. They get you into the sources AI reads. But the new work is brand: being the answer it reaches for first.
