News & analysis
AI Tools Now Rank Brand Recommendations, Not Just Websites
A new framework argues that ranking well in Google says less than it used to, because AI tools now decide who gets recommended and who gets left out of the answer entirely.
By Roomer Editorial · · 4 min read
What's the difference between a search ranking and an AI answer?
A ranking is a position in a list. The person doing the search still has to scan the options and pick one.
An AI answer works differently. The model has already compared the options before the user reads anything. The source frames it this way: you don't lose the click so much as discover you were never in the running for it.
That matters for anyone whose content shows up when someone asks an AI tool "best boutique hotel in Porto" or "where to eat in Oaxaca city." The click isn't just harder to win. For a growing share of queries, it stops being available at all.
What are the three gates agencies should be measuring?
The source lays out three distinct things an AI answer can do, and argues most reporting conflates them.
Inclusion is whether you show up at all — a brand mention or a cited source. Useful, but not proof of anything beyond a seat in the conversation. A model can cite your article to explain a problem and then recommend a competitor to solve it.
Framing is what the model says you are. The source gives the example of a B2B software company that expanded into enterprise customers but still has old product pages describing it as a tool for freelancers — the AI tool repeats that outdated framing to an enterprise buyer, and the company loses relevance before anyone reaches its site.
Endorsement is whether you're the actual pick on a decision-stage question. The source calls this "the gate worth fighting over," since a brand can clear inclusion and framing and still lose the recommendation to a competitor the model judges better suited.
What do the numbers actually show?
The source cites a Similarweb study (opted-in US desktop panel, July to December 2025) on what happens after ChatGPT recommends a brand. When ChatGPT recommended American Express, 7.2% of users went on to visit its site, versus 3.1% who visited Capital One. When Capital One got the recommendation, 14.2% visited its site versus 3.8% for American Express.
The study covered large consumer brands, and the source is explicit that the findings may not hold for smaller or B2B businesses.
On raw traffic, the source points to Ahrefs' opt-in Web Analytics panel, where AI assistants made up under 1% of site visits in September 2026. That figure reflects what participating sites actually received, not how often an AI tool hands off a click in general.
The source's broader point: if AI endorsement share rises alongside branded search and leads, that's worth investigating, not treating as proven cause and effect.
Roomer Editorial's take
The source is written for agencies and B2B software brands, but the underlying shift applies directly to travel content. If someone asks an AI tool where to stay in Lisbon, the model is pulling from whatever it can crawl and compare — and a video that lives only inside a social app isn't part of that comparison. A page with structured content and a bookable link is.
This is the argument for having your own site in the first place, not as a vanity project but as the thing an AI tool can actually cite. Roomer gives creators a travel site built from their own videos, with bookable pages and booking rails through Vio, Stay22, GetYourGuide, Viator and Klook — inventory a model can point to when it answers a question about where to book, not just where to look.
The inclusion gate matters most here: if your content about a destination lives only inside a platform's algorithm, you're not even in the set of things being compared. Getting a page indexed is table stakes before framing or endorsement are even relevant.
Our view, and it's an inference the source doesn't make: creators who treat their own site as the canonical version of their recommendations — reviews, hotel picks, tour bookings, all in one place — are the ones with a shot at being cited accurately when a reader asks an AI tool instead of Google. Creators who only post to social lose that chance by default, not because their content is worse, but because there's nothing for the model to point to.
The exception: this doesn't mean rankings stop mattering. Organic search is still, per the source, "the traditional baseline" running alongside AI answers. Don't abandon SEO basics to chase AI citations you can't yet measure reliably.
Roomer Editorial's view: start with the basics before worrying about AI visibility at all. Make sure your site actually describes what you cover accurately — destinations, price range, the kind of traveler you're writing for. An AI tool can only repeat what's already on the page, so an outdated bio or a stale "about" section will get echoed back to anyone asking for a recommendation. Fix that first, because it's the one thing on this list you can control directly, today, without waiting on any platform to change.
One action: audit what an AI tool says about your top three recommended properties or tours right now, and check whether it matches what's actually on your own site.
Source: Search Engine Journal
