What changed in travel search
For two decades, travellers planning a trip opened Google. They typed a query, scanned ten blue links, opened five tabs, and built a shortlist. Operators that ranked in the top three got the click; operators in positions four to ten got partial attention; operators below page one got nothing. The whole industry — SEO, paid media, content marketing, OTA distribution — was built around that ranking surface.
That surface is breaking down.
Travellers now ask ChatGPT for “small-group hiking tours in the Scottish Highlands,” Perplexity for “boutique lodges in Cornwall with a sea view, mid-range,” Claude for “best Sunday lunch in Bristol with vegetarian options,” and Google’s AI Overviews for “where to stay near Buttermere with a dog and a fire.” The answer is rarely a list of ten links. It’s a synthesised paragraph naming three to five operators directly, with a one-line description of each, and citations to the sources the model pulled from.
If your business is one of the named operators, you’ve won the discovery round before the traveller has even opened a tab. If you’re not, you’ve lost it before you knew the conversation was happening.
This isn’t a minor channel shift. Across the 240 properties Booked Wild audited in September 2025, 87% were functionally invisible to ChatGPT for buyer-intent queries in their region. The model returned the OTA listing — Booking.com, Tripadvisor, GetYourGuide — and the operator’s own brand and domain went unmentioned. The OTA wins the citation, owns the click, and takes the commission.
For independent travel and hospitality operators across Europe, that 87% figure is the new strategic problem. AI visibility isn’t an additional acquisition channel; it’s becoming the precondition for being considered at all.
What AI engines actually read
AI search isn’t Google with a chatbot bolted on. It’s a different evaluation surface, with different signals, that produces a different output shape. To understand what to fix, it helps to understand what the model is reading when it composes an answer.
Three categories of signal matter most:
On-site structured signals. Schema.org JSON-LD markup (LodgingBusiness, TouristAttraction, Tour, LocalBusiness, FAQPage), llms.txt and AGENTS.md at the site root, plain-text pricing on a public URL with Offer schema, citation blocks formatted as 50–80 word direct answers, and FAQPage markup on conversion pages. These aren’t ranking signals in the traditional SEO sense — they’re extraction signals. They tell the model what your business is, what it offers, what it costs, and what questions it answers, in a format the model can parse and quote without hallucinating.
Off-site authority signals. Tripadvisor profile completeness and review velocity, Google Business Profile categorisation and post cadence, third-party editorial mentions on sources AI engines weight (regional travel publications, niche aggregators like Sawday’s, Sidetracked, Cumbria Life, Daily Finland — the editorial sites that have themselves been cited in AI training data). These signals corroborate what your own site claims. AI engines prefer sources that agree with each other; an operator whose own site, Tripadvisor, GBP, and a Cumbria Life article all describe them the same way is a high-confidence source. An operator with a beautifully written about-us page and an empty Tripadvisor profile is a low-confidence source.
Engine-specific behaviours. ChatGPT and Perplexity disagree. Across 4,800 prompts in the September 2025 research, citation overlap between ChatGPT-4o and Perplexity on identical queries was 41%. Claude is the kindest reader — 18 percentage points more likely to surface independent operators over OTAs for the same prompts. Google AI Overviews lean heavily on Google Business Profile signals and review velocity. Each engine reads slightly differently, which means optimising for one doesn’t fully optimise for all — but the structural work that lifts ChatGPT typically lifts the others too, just by different amounts.
The cumulative implication is straightforward. The signals that matter aren’t keyword density, backlink count, or content volume. They’re machine-readability, structural clarity, third-party corroboration, and named human authority. The operators winning AI citations in 2026 are the ones whose websites, listings, and editorial presence converge into a single high-confidence answer.
The four-layer fix
There’s a tendency to treat AI visibility as a single project — “do the AI thing.” It isn’t. It’s four layers of work, each compounding on the others, each addressable independently. We sequence the work this way because the structural foundation matters before the editorial layer; building third-party citations before your own site has clean schema is like running paid media to a slow checkout. The mechanics matter.
Layer 1: On-site structured data
The single biggest lever, and the cheapest. Deploy LodgingBusiness, Tour, FAQPage, Service, Article, and BreadcrumbList schema across your site. Validate every page against Google’s Rich Results Test. Publish llms.txt and AGENTS.md at the site root. Surface your pricing as plain HTML text with Offer schema attached.
Most operator sites we audit have some schema — a generic Organization block someone added in 2019, often broken — and almost no FAQPage or Tour schema where it would actually move the needle. The fix is two weeks of work for an operator with a typical 10–30 page site. Median time from schema deployment to first AI citation in our cohort data: 11 days.
Layer 2: Tripadvisor and Google Business Profile
The two off-site profiles AI engines read first. Tripadvisor remains the dominant source for AI restaurant and lodging citations — the platform we’ve been told is dying isn’t, it’s the source of truth for the model. GBP categorisation and review velocity decide map-pack queries (“best Italian in Bristol,” “lodge near Buttermere with a hot tub”). Profile completeness, structured Q&A, photo metadata, and response cadence on both platforms compound into the citation weight that AI engines pull.
Most independent operators have weak versions of both — generic Tripadvisor descriptions, empty Q&A, GBP profiles missing core attributes (dog policy, EV charging, family rooms, specific cuisines). The £395 Tripadvisor + GBP service rebuilds both in a week. Citation lift is typically visible within 2–4 weeks.
Layer 3: Plain-text pricing and named-author content
If your pricing lives in a PDF, an image, a third-party booking widget, or a hidden modal, AI engines can’t read it. Worse, they often hallucinate “pricing varies — contact for details,” which is a lost enquiry every time. Surface every price as plain HTML text on a public URL with Offer schema attached. Same applies to named-author content — AI engines reward content with named, credentialled human authors over anonymous “by the team” pages. Pages with Person schema attached to a real bylined author cite measurably better.
This sits inside the £495 AI Visibility Fix, but it’s worth calling out separately because the impact is so out of proportion to the work. A weekend’s content reorganisation routinely shifts citation rates more than three months of paid media.
Layer 4: Third-party editorial citations
The slowest layer to build, the most defensible once it’s in place. AI engines reward operators mentioned in editorial context on third-party sources they trust. For UK operators that’s titles like Cumbria Life, Sidetracked, Trail, Sawday’s, the regional press in your county, and niche aggregators in your sector. For European operators it’s the equivalent regional and specialist publications in each market. Three to five named editorial mentions on sources AI engines weight is enough to shift citation share materially; ten to fifteen makes you the default answer for your category.
This layer doesn’t substitute for the first three — citations to a site with broken schema and an empty Tripadvisor profile won’t lift you. But layered on top of clean structural work, editorial citations are what move you from “occasionally cited” to “named in the answer most of the time.”
What it costs to do this
We post our prices for the same reason we recommend you post yours — trust is a function of legibility. The work above is delivered through three productised services:
- £395 — Tripadvisor + GBP Optimisation. One week. Layer 2.
- £495 — AI Visibility Fix. Two weeks. Layers 1, 2 (lighter touch), and 3.
- £97 — Deep Insight Audit. Two weeks. Document only — diagnostic across all four layers, with prioritised next steps. We recommend this as the first step for operators not sure which fix they need.
Most operators do the £97 audit, then layer in the £395 and £495 services in sequence over three months. Layer 4 (editorial citations) is built more slowly, often as part of a longer engagement. Total first-year investment for the structural work sits in the £1,000–2,000 range for most independents, with the £4,500–7,500 AI-Ready Website Rebuild being the right answer when the existing site is the structural bottleneck.
What this isn’t
A guarantee. AI engines have no fixed ranking surface, citation share rises and falls with model retraining and competitor activity, and no honest agency promises specific citations on specific dates. What we guarantee is the structural work, the prompt re-test, and the day-30 follow-up audit. If the same prompts show no movement at day 30 on the £495 fix, we run another round of intervention at no charge.
A short-term tactic. The signals AI engines reward — schema, citations, named authority, structural clarity — are the same signals classical SEO has wanted for a decade. They compound over years; they don’t decay in a quarter. The operators investing in the structural layer in 2026 will benefit from compounding citation patterns through 2027–2030. The ones who delay will face calcified competitive positions where competitors own the recommendation slots.
A substitute for craft. AI citation share is a discovery layer, not a delivery layer. Operators that get cited but deliver an average experience will lose repeat bookings the same way they always have. The work above gets the right traveller through the door; what they tell their friends afterwards is still your job.
The window
AI-mediated travel discovery has accelerated through 2025 and 2026. Travellers are increasingly asking AI engines for recommendations rather than browsing OTAs or running Google searches. The structural shift is already underway, and it’s compounding because AI engines develop persistence bias — once a brand is consistently cited, that pattern reinforces.
The window for independent operators to build AI citation share is open now. Operators who establish AI visibility through 2026 will benefit from compounding citation patterns; operators who delay will face calcified competitive positions where competitors own the recommendation slots in their region.
The four layers above are the structural work. The £97 audit is where most operators start. If you’d rather just see what we’d find on your business specifically, that’s the next step.