SEO is dead. Long live AEO.

The most consequential change to restaurant marketing in 2026 isn’t anything anyone is marketing. It’s what’s happening to Google.

A growing share of restaurant queries don’t return ten blue links anymore. They return a single answer, generated by a language model, citing two or three sources by name. Google AI Overviews, Gemini, ChatGPT Search, Perplexity, Claude: different products, same shape. A question goes in, an answer comes out, with sources footnoted below.

Ask Perplexity right now for the best sourdough in central Singapore and you’ll get a short paragraph naming three bakeries, with citations to a food blog, a Reddit thread, and an Instagram post that ranked them. None of the cited sources paid to be there. None had a marketing agency in the loop. What they had was a named human writing carefully about a specific thing, on a platform the engine indexes. That’s the new mechanic.

A new acronym is already in circulation for the discipline: AEO, for Answer Engine Optimisation. Answer Engine Optimisation is the practice of structuring content so that AI synthesis engines (ChatGPT, Gemini, Perplexity, Claude, Google’s AI Overviews) cite it when answering a user’s question.

For venues, this changes the economics of being findable. You’re no longer optimising to land on Page 1. You’re optimising to be one of the names the engine cites when someone asks it about your kind of place. And the interesting thing, the reason this piece is on this site, is that almost everything AEO rewards is what a tastemaker network produces by default. The shift from SEO to AEO is, mechanically, a shift in favour of the kind of marketing that already worked for small opinionated venues.

The network wasn’t built with AI search in mind. It was built around taste, and now finds itself unintentionally in front of the search shift.

What actually changed

The shortest way to put it: search engines used to be matching engines, and now they’re synthesis engines.

Old Google read your query as keywords, matched them against indexed pages, and returned a list of links for you to choose from. The optimisation game was to rank in the top three blue links for the keywords your audience searched.

Modern AI-augmented search reads your query as a question, decides what a good answer looks like, builds the answer from multiple sources, and returns the synthesised paragraph with citations to where each claim came from. The optimisation game is to be one of the citations.

This sounds incremental until you notice that the user often doesn’t click anything anymore. They already got their answer. Click-through rates from AI Overviews are reportedly down 30 to 50 per cent compared with traditional results. The traffic that used to flow to whoever ranked first now flows to whoever was cited, and a meaningful chunk of it doesn’t flow anywhere at all.

For small businesses, this is a serious problem if your traffic model assumed “rank well, get the click.” It’s an opportunity if you can become one of the things the AI cites.

Search engines used to be matching engines. Now they’re synthesis engines. The optimisation game is to be one of the citations.

What AI search rewards

What does the engine use to pick its citations? Four things, in plain English. The official acronym, E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness), covers the first two; the rest is a vocabulary problem the field hasn’t agreed on yet.

Lived experience.

First-hand experience is now the top-weighted citation signal for AI search. Content based on actually doing the thing (going to the place, eating the meal, talking to the staff) gets preferred over content that is clearly aggregating other people’s reporting. The engines can now tell the difference reliably and the gap is widening. We’ve written a longer field guide on what this looks like on a feed in How to find credible food creators.

Named humans.

A real byline with a real biography matters more than it used to. Engines are being trained to distinguish between content with editorial provenance and content that’s anonymous or visibly AI-generated. Sites with a history of being right, written by named people whose other work is verifiable, get cited; ghost-written marketing pages don’t.

Distributed mentions.

The new backlink. The engine notices when a venue or person is mentioned across multiple respected sources by name, in context, on different platforms. Each mention is a citation candidate; the cumulative pattern is a reputation signal that outweighs any single piece of SEO content, and that no amount of agency-driven content can fake.

Machine-readable hygiene.

Schema markup, clean headings, FAQ-shaped answers, anchor IDs, descriptive alt text. The boring stuff that tells the engine what each piece of text is and lets it cite a specific paragraph rather than a whole page. People talk to AI engines the way they talk to a friend, so the content that gets cited is content that already answers the question in natural-language form.

Why tastemaker work produces all of this by default

Now apply that list to what actually happens when a tastemaker visits a restaurant and posts about it.

A named human (the tastemaker) writes about their first-hand experience (the visit) on a platform that already carries their expertise credentials (the previous posts in their feed, their established voice). The post lives on a third-party platform, not on the venue’s own site, so it functions as an independent mention. That’s the kind of signal engines weight as external trust.

Run six tastemaker placements over six months and you have six independent first-hand reports from credible named humans, on three different platforms, in conversational long-tail form, each one a piece of the topic cluster around your venue. The cumulative signal to any synthesis engine is: this place exists, multiple credible people have been there, here is what they specifically observed.

Now compare that to the traditional alternative. A paid influencer placement, the kind we’ve spent a fair bit of this site arguing against, produces one post on one platform, marked sponsored, often visually templated, by a creator whose feed is half rate-card. As an AEO signal it’s actively negative: engines specifically discount sponsored content because it’s not credible first-hand reporting. The post may rack up reach. The synthesis engine won’t quote from it.

That’s what we mean when we say a tastemaker network is well-suited to AEO. The model wasn’t built for AI search; it was built for credible-voice marketing, and the two turned out to want the same things. The longer argument for credible-voice marketing lives in Influencer marketing has failed.

A paid placement may rack up reach. The synthesis engine won’t quote from it.

What this means for venues

If you’re planning your 2026 and 2027 marketing around traditional SEO content and paid influencer placements, you’re optimising for a search world that’s already shrinking. The kinds of marketing that will keep working as AEO takes over are different in shape from what most agencies still sell:

• A small but credible network of voices posting about your venue, over time, from real first-hand experience, on platforms AI engines crawl heavily.
• Coverage in respected publications (local food press, regional travel writers, neighbourhood newsletters) where you’re mentioned by name with context, not in a paid placement.
• A clean site of your own with structured data, a real about page with a named author, an FAQ that answers the questions your customers actually ask.

The third is housekeeping you can do in a weekend. The first two are what a tastemaker network does (the operating details are in how the network works) and what good restaurant PR has always done. The cost of skipping them is that nothing surfaces when an AI engine tries to answer a question your venue should be the answer to.

What this means for tastemakers

If you write about food and drink in your own voice, to a small audience, with first-hand specificity, your work has just become meaningfully more valuable, even if your follower count hasn’t moved.

Five years ago, your post about a Tiong Bahru bakery reached a few hundred people who already followed you. Today, that same post might be the source Gemini cites when someone in another city asks where to get a good sourdough in central Singapore. Your audience is suddenly larger than your follower count, because AI synthesis engines are pulling from your work as primary source material.

This is also why the bar to get into the network is what it is. The signal engines reward is real first-hand observation by a named human with editorial credibility. Generic “OBSESSED with this brunch spot” posts don’t make the cut. Specific, observed, written-by-you posts do. The criteria we read every application against live in What we look for in a tastemaker.

Three things to do

For any venue reading this, three small moves, in increasing order of how much they’ll bend your year:

1. Clean schema markup on your own site (LocalBusiness, FAQPage, Article), photos with descriptive alt text, an about page that names a human.
2. A named editorial author for any content you publish, with a verifiable biography. Not a stock “Marketing Team” attribution.
3. A small standing relationship with three to five credible voices in your scene. Yes, that’s what we sell. We would argue for it even if we didn’t, because the alternative for a small venue is to write everything yourself and hope the engine notices.

The shift from SEO to AEO is the first marketing turn in a decade where being small and credible isn’t a handicap. It’s the actual shape of what the engines now want. The venues being cited next year are the ones who start acting like publishers this year.