If a hungry local in your area opens ChatGPT right now and types "best halal kebab in Sheffield" or "good biryani near me in Bradford", what does the AI reply with — and does it name your restaurant?
For nearly every independent UK takeaway we audit, the answer is no, your restaurant isn't mentioned. The AI reply lists two or three competitors — usually the loudest ones, not the best — and you don't get a look in. The diner closes ChatGPT and orders from whoever the AI named.
This isn't a "Google search results" problem. It's a different problem. And it's growing fast, because more people every month ask ChatGPT, Perplexity or Google's new AI Overviews before they ever open a maps app.
This guide walks through what's actually happening, in plain English. Read it whether you want to do this yourself or just be a smarter buyer if you decide to outsource it.
What "AI Search" actually means
"AI Search" is the umbrella term for getting cited inside an AI engine's reply. The disciplines have a few names, all describing roughly the same work:
- AEO — Answer Engine Optimisation
- GEO — Generative Engine Optimisation
- LLM SEO — making your site readable by Large Language Models
We don't really use the jargon with clients. The work is the same either way: rebuild your site so the AI engines crawling it can pull clean facts about your restaurant — the menu, the hours, the dietary tags, the area you serve — and feel confident enough to cite you in their reply.
The engines that matter, today, in the UK:
- ChatGPT (with search) — OpenAI launched search-in-ChatGPT in late 2024[1]
- Google AI Overviews — the AI-written paragraph at the top of Google results[2]
- Perplexity — increasingly used for local "find me…" queries
- Gemini — Google's standalone AI assistant
- Apple Intelligence — quietly answering more queries via Siri and Spotlight
Each of these reads your site slightly differently. The good news: optimising for one mostly helps with the others.
Why AI Search matters for takeaways specifically
Restaurant discovery is one of the most AI-friendly use cases there is.
Think about how people search for food:
- "best vegan pizza in Manchester"
- "halal biryani near me that's open right now"
- "anywhere doing gluten-free pakora?"
- "good cheap kebab walking distance from BD1"
These are natural-language questions. AI engines were built to answer natural-language questions. Traditional search engines were built to match keywords. The gap between what people type and what they want has never been bigger — and AI has narrowed that gap in a way Google's blue links never did.
The diner who used to open three apps and compare star ratings is now asking one question and trusting the reply. If you're not in the reply, you don't exist.
How AI engines decide who to cite
Three signals matter most.
1. Structured data they can actually read
When an AI crawler visits your site, it doesn't read your beautiful homepage hero. It reads the structured data behind the scenes — JSON-LD Restaurant, Menu and MenuItem schema that tells machines exactly what's on your menu, what dietary tags apply, what your opening hours are and which area you serve.[3]
Almost no independent UK takeaway has this in place. The big chains do. That's why AI engines keep citing chains for queries that would naturally favour local independents.
The fix: a proper schema rebuild. Menu items tagged with suitableForDiet: HalalDiet (or VeganDiet, GlutenFreeDiet), hours wired into openingHoursSpecification, geo coordinates so the AI can map your restaurant against "near me" queries.
2. A clean /llms.txt manifest
In 2024, Answer.AI proposed a new file standard called llms.txt — a simple markdown file at the root of your site that tells AI crawlers what's on the site and how to interpret it.[4]
It's the AI equivalent of a sitemap, but specifically designed for how LLMs prefer to ingest content. ChatGPT, Claude and Perplexity all consult it when it exists.
Most takeaway sites don't have one. Adding one takes an afternoon for a developer and can be the difference between an AI engine quietly skipping your site and confidently citing it.
3. Direct-answer paragraphs
AI engines extract answers from sites by lifting paragraphs almost verbatim. The paragraphs they prefer are direct-answer format: 40–60 words that answer a specific question without preamble.
So instead of writing:
Our restaurant has been serving the local community since 2010 and we take pride in our authentic dishes prepared by our skilled chefs who use only the finest ingredients sourced from local suppliers...
You write:
Is your biryani halal? Yes — all our chicken, lamb and beef biryanis are prepared with halal-certified meat from suppliers in the West Midlands. We also offer a vegan biryani made with grilled paneer-substitute and seasonal vegetables. Our kitchen is fully separated from non-halal preparation.
The second version is what AI cites. The first version is what AI skips.
What to do this month (DIY version)
If you're going to attempt this yourself, here's the priority order.
Week 1 — see how invisible you currently are.
Open ChatGPT, Perplexity and Google AI Overviews. Ask each one your top five most-important queries — "best [cuisine] in [your area]", "halal [dish] near [postcode]", "[cuisine] open late in [city]". Screenshot what you see. Note which competitors get named.
This is your baseline. You can't improve what you don't measure.
Week 2 — fix the menu data.
Get your menu into a structured format. If your site is on a modern platform (Next.js, Webflow, modern WordPress), this means adding Menu and MenuItem JSON-LD schema for every dish, with prices, descriptions and dietary tags. If you're on Just Eat or Deliveroo only, this becomes much harder — you don't control the page AI is reading.
Week 3 — write the answer blocks.
Add a FAQ section to your site with 6–10 of the most common questions diners ask about your kitchen. Write each answer as a 40–60 word paragraph that directly answers the question. Wire it up with FAQPage schema so AI engines recognise it.
Week 4 — publish the llms.txt.
A developer can add a /llms.txt route to your site in an afternoon. It lists your services, your menu, your locations and your blog posts in a clean markdown format. Test it by visiting yourdomain.co.uk/llms.txt — it should return a plain text file with all your key facts.
What success looks like
You're not aiming to be cited by every AI engine for every query overnight. You're aiming for these milestones, in this order:
- Month 2 — your site appears in Bing search (which feeds ChatGPT search) for at least one brand-related query.
- Month 3 — you're cited by one of the four big AI engines for one of your priority queries.
- Month 6 — you're cited by multiple AI engines for multiple queries, and you've started outranking at least one of your initial-baseline competitors.
- Month 12 — you're the default citation for your cuisine in your area, and you can demonstrate it with side-by-side screenshots from your baseline.
Most takeaways that put the work in see their first AI citations between weeks 4 and 8.
What success doesn't look like
A few honest caveats so you don't waste money:
- No guarantees. AI engines change their citation behaviour every few months. Anyone promising "guaranteed ChatGPT placement" is selling you something they can't deliver.
- AI Search is still smaller than Google Maps. For most takeaways, the majority of new customers still come from Google Maps, Just Eat and word-of-mouth. AI Search is the fastest-growing channel, not the biggest one.
- It compounds slowly, then suddenly. You'll see nothing for the first few weeks while AI engines re-crawl your site. Then citations start landing in clusters.
- Reviews still matter. AI engines weight your visibility partly on the strength of your reviews. If you're sitting on 12 reviews with a 3.8 average, AI is going to be reluctant to cite you over a competitor sitting on 200 with a 4.6.[5]
The short version
AI Search is the work that gets your restaurant named when locals ask AI engines instead of Google. Three things matter: schema, llms.txt, and direct-answer content. Most takeaways have none of this. The first ten citations are cheap to win.
If you want help, Manto Studio's AI Search service is included in every package — Grow gets the basics in place, Dominate adds active optimisation, Scale adds quarterly mention tracking. Or if you'd rather try it yourself, this article is the playbook — work through the weekly plan above and you'll see your first citations land before the end of next month.
Sources & further reading
We help UK restaurants and takeaways grow online. Get in touch for a free audit.