Case study
How Fozias restaurant went from invisible to #1 on Perplexity in 90 days
Fozias is a Pakistani and Kashmiri restaurant in Liverpool — halal, alcohol-free, and family-run. In January 2026 its AI visibility score on Visus was 23 out of 100: strong on Google reviews, but nearly invisible to generative engines. Ninety days later the same measurement framework showed 80/100. Today, for queries like “Kashmiri restaurant Liverpool,” Perplexity cites Fozias as the top recommendation. This case study walks through what was broken, what we changed week by week, and what it means if you run a local UK business.
The starting point
Fozias serves authentic cuisine from 35 Renshaw Street, inside the Grand Central building in Liverpool city centre. The dining room is busy, repeat customers leave five-star Google reviews, and the brand has real word-of-mouth — exactly the profile that should surface when someone asks an AI for “halal restaurant Liverpool” or “Kashmiri food near me.”
Yet the first audit told a different story. The site had no meaningful schema, no llms.txt, and content written for humans in marketing tone rather than extractable facts AI systems can quote. There were no sameAs links tying the website entity to Instagram, Facebook, or trusted directories. Schema, where it existed, was a generic LocalBusiness blob — not the specific Restaurant type with cuisine, hours, and review objects that models use to disambiguate “a place to eat” from “a company with an address.”
When we probed Gemini with the live domain, the entity state for fozias.co.uk came back “Unknown” — the digital equivalent of the model having no reliable file on the business.
That single signal mattered more than vanity keywords ever could: if the model does not know what you are, it cannot recommend you with confidence.
What we changed
Week 1 — Schema markup
We replaced the generic local entity markup with a full Restaurant schema graph: servesCuisine (Pakistani, Kashmiri), structured opening hours, geo coordinates for the Grand Central location, and amenities that match how people search — including halal, alcohol-free, and family-friendly signals where accurate.
We added a real AggregateRating aligned with live Google data — 4.8 stars across 127 reviews at the time of implementation — and linked out to third-party proof (e.g. coverage in YM Liverpool, Liverpool Echo) so corroboration did not rest on the site alone.
Week 2 — Content structure
We rewrote key blocks so every sentence carried something a model could verify or quote. Instead of “passionate team” and “amazing experience,” the homepage and About sections stated facts: Fozias is Liverpool’s dedicated Kashmiri restaurant, serving authentic halal cuisine at The Grand Central on Renshaw Street, with dine-in and takeaway, open seven days. That pattern — who, what, where, when — is what large language models extract when they build entity summaries.
Week 3 — Entity building
We added sameAs URLs to Instagram, Facebook, and major UK listings, published an llms.txt at the site root summarising the business and linking priority pages, and audited NAP consistency (name, address, phone) so the schema, visible HTML, and footer matched character-for-character where possible.
Week 4 — Citation building
We completed profiles on Yell, Yelp UK, and TripAdvisor, and fully optimised Google Business Profile: correct primary category, service areas, attributes (halal, family-friendly), and 10+ high-quality photos. The goal was not “more links” in an SEO-spam sense — it was consistent corroboration that Perplexity and others could cross-check.
The results
- Perplexity cites Fozias first for high-intent local queries in this niche.
- Gemini moved from “Unknown” to “Known” for the domain entity check.
- ChatGPT began including Fozias in curated lists for halal and regional cuisine prompts where Liverpool was in scope.
- The Visus score climbed from 23 to 80 over roughly ninety days — not overnight, but fast enough to matter in a competitive city centre.
The single highest-impact change was moving from vague LocalBusiness JSON-LD to a specific Restaurant schema with real AggregateRating and press links — worth about 12 points on our internal model before any off-site work compounded.
What this means for your business
AI search is not a 2027 problem. It is happening on phones and desktops today: hundreds of millions of weekly ChatGPT users, hundreds of millions of monthly Perplexity queries, and AI Overviews on Google. The businesses that treat GEO as seriously as SEO — schema, entity clarity, citations — are earning shelf space in those answers.
The encouraging part for independents: most of this is technical and editorial work, not a seven-figure ad buy. Ninety days is enough time to fix baseline schema, rewrite thin copy, add llms.txt, and tighten listings — if you know exactly what engines reward. That is what Visus is built to surface.
What’s your AI visibility score?
Find out in 60 seconds. Free — no account required.
Check your score freeGet this done for your business
Sirat Media implements AI visibility programmes for UK businesses — schema, entity fixes, llms.txt, monitoring, and ongoing citation hygiene — with Visus included for measurement. Engagements from £125/month, no setup fee. Learn more at emicomedia.co.uk.