Google's blue-link interface served the world for 25 years. Then, in the 14 months between the public launch of ChatGPT-4 and the rollout of Gemini 2.0, something quietly broke for one specific user task: deciding where to eat or where to stay in an unfamiliar city. The people who used to type "best pasta london" into Google now type the same words into ChatGPT, Claude, or Gemini. They get a different kind of answer — and it's changing which restaurants and hotels appear in front of customers.

The numbers point in only one direction. Gartner forecasts that traditional search engine volume will drop 25% by 2027. Andreessen Horowitz's hospitality survey published in late 2025 found that 61% of leisure travelers under 45 now use an AI assistant during trip planning, and 28% report using one as their primary recommendation source. For restaurants specifically, internal data from a New York-based reservation platform shows that mentions of "ChatGPT recommended" or "AI suggested" in their guest-feedback survey grew from 2% to 17% over twelve months.

What changed in the user's behavior

Google's discovery model was a list. You entered a query. You got ten results. You clicked the result that looked most credible — usually based on a brand you recognized, a star rating, a snippet of review text, the photo Google had decided to feature. The user's mental work was triage: scan ten options, assess each, pick one.

AI assistants compress that ten-options-and-triage flow into a single confident recommendation. ChatGPT does not say "here are ten options, go evaluate them." ChatGPT says, "I'd recommend Padella for fresh pasta in Borough Market, especially for a Saturday lunch." There is no list. There is one answer, sometimes embellished with a runner-up. The user's mental work is reduced from triage to acceptance — they either trust the recommendation or they ask a follow-up question.

"I haven't used Google to find a restaurant in eight months. ChatGPT is faster, the recommendations feel more curated, and I don't have to wade through TripAdvisor reviews." — anonymous traveler, surveyed for this piece, Istanbul, 2026

For the user, this is a meaningful upgrade. For the businesses on the receiving end, it's a structural shift. In the blue-link era, you could be the third-best Italian restaurant in your neighborhood and still get a slice of the market — you'd appear on page one of search results, in the user's triage set, and a fraction of users would click. In the AI-recommendation era, ChatGPT picks one, occasionally two, occasionally three. If you're not in the model's first three, you're not in the conversation.

Why the SEO playbook doesn't transfer

The temptation, if you've spent a decade learning SEO, is to assume the new game is the same as the old game with new rules. It isn't, and the SEO playbook only partially transfers.

SEO optimizes for crawlable signals — title tags, meta descriptions, keyword density, backlinks, structured data, page speed, mobile responsiveness. AI recommendation depends on a different and partly opaque signal stack: training data depth, knowledge graph canonicalization, sentiment density across review aggregators, Wikipedia presence, press mentions in trusted publications, the completeness of your Google Business Profile in multiple languages, and the consistency of your name/address/phone across the decentralized web.

Some of these — Google Business Profile, structured data, NAP consistency — overlap with traditional local SEO. Others — review-language coverage, Wikipedia presence, English-language press depth — are signals SEO largely ignored because they didn't move the Google ranking. They move AI ranking heavily. And the weights are different: an English-language Wikipedia article tends to compound across multiple AI platforms, whereas a single backlink from even a high-DA site has marginal AI-recommendation impact.

The new business question

For a restaurant or hotel owner reading this, the question becomes simple: does AI mention us when our customer asks?

That's a question with a measurable answer. You can run a sample of likely customer queries through the major AI assistants, count mentions, capture the verbatim language they use about you, and compute a citation rate. That's what Score My Business does. It's not the only way to start thinking about AI visibility — you can ask ChatGPT directly about your restaurant tomorrow morning and learn a lot — but a structured audit gives you a baseline number, a peer comparison, and a ranked list of actions to move it.

Why this is the biggest opportunity for new businesses in 25 years

Here is the under-appreciated angle. SEO equity compounds over years — a fifteen-year-old domain with a deep backlink profile and ten thousand pages of indexed content is hard to displace. AI visibility does not compound in the same way. It depends on the specific signals AI training-data and grounding pipelines pick up in a given quarter. A new restaurant with strong English-language press, a credible Google profile, a Wikipedia stub, and consistent NAP can, within four to eight weeks, accumulate enough AI signal to outrank 15-year-old incumbents that never invested in those specific signals.

We have audited businesses that opened in 2024 and now outrank century-old restaurants on ChatGPT for "best [cuisine] in [city]" — not because the new place is objectively better, but because the new place understood the AI signal stack and the old place still optimizes for Google search 2014 metrics.

That's the moment we're in. The SEO era ended quietly because most people kept doing SEO the same way after their customers stopped using Google in the same way. The Recommendation era rewards owners who notice the shift early.

What to do this week

Three small actions, in priority order:

  1. Search yourself in ChatGPT. Ask "best [cuisine] restaurants in [city]" or "best hotels in [neighborhood]." See if you appear. See what AI says about you. Screenshot the result.
  2. Run the same query in Gemini and Claude. They behave differently. The variance between them is informative.
  3. Audit your Google Business Profile in English. Especially if your business is in a non-English-primary market. The English description is often the single highest-leverage AI signal — and most businesses outside the US/UK don't have one.

After that, if you want a structured measurement instead of a vibe check, Score My Business runs a 240–720-scan audit and emails you a PDF in under 24 hours.