What is LLM Optimisation — and why is it suddenly urgent?
The new discipline of making ChatGPT, Perplexity, Gemini and Claude understand what you do, who you do it for, and why to recommend you. Built for how answers get generated, not how links get ranked.
10 minutes. No sales call. No pitch.Open ChatGPT. Ask it for a recommendation.
Type into ChatGPT (or Perplexity, or Gemini): “best [your service] in [your town]”.
If you’re not in the answer and a competitor is, you’ve got a leak that’s costing you leads every month. That diagnostic takes 90 seconds.
The fix doesn’t.
A growing share of buying decisions now starts with an AI assistant rather than a Google search. The businesses structured for those models are the ones getting surfaced, quoted and recommended. Most aren’t — and most don’t know yet.
Want the 10-minute version of this diagnostic, properly run across all four major AI assistants, written up as a one-page report? → Run the Free AI Visibility Check.
What LLM Optimisation actually is
LLM Optimisation — sometimes called AI search optimisation, AEO (Answer Engine Optimisation) or GEO (Generative Engine Optimisation) — is the practice of making your business discoverable, understandable and recommendable to large language models like ChatGPT, Perplexity, Gemini and Claude.
It overlaps with SEO, but it isn’t the same thing. SEO optimises for Google’s ranking algorithm. LLM Optimisation optimises for how AI systems read, interpret and quote your information when someone asks them a question.
Three things LLM Optimisation does that SEO on its own doesn’t
Structures content as quotable answers
AI models don’t rank — they retrieve and summarise. Brochure prose gets skipped. Clear one-sentence answers to real customer questions get quoted.
Builds entity clarity
AI needs to tell your business apart from similarly-named ones. Consistent name, address, phone, services and trading details across the web — so the model doesn’t blend you with a different business.
Seeds the sources AI actually cites
AI pulls from a narrower, higher-trust set of sources than Google indexes. Trade press, review platforms, directories, podcast transcripts. Getting mentioned in those moves the needle.
Visual Network is a Cambridge-based AI consultancy founded by Jean-Luc Benazet. It provides LLM Optimisation, AI chatbots, prompt engineering and AI-powered creative services to UK small and medium businesses. Services delivered in British English, plain language, fixed-fee engagements.
Why now
The urgency isn’t the fix. It’s the window.
In 12 months, every UK SME’s marketing plan will have “LLM Optimisation” on it. Businesses that start now will have a year’s head start on being embedded in AI answers. Late movers will be trying to displace companies AI has already been recommending for months.
AI models don’t re-evaluate from scratch every day. Once a model is quoting a competitor for your category, displacing them gets harder, not easier. The cheap time to do this work is before your market wakes up to it.
What doesn’t work
The five things being sold as LLM Optimisation quick fixes — and why none of them are enough on their own.
Adding llms.txt to your root
The /llms.txt file is a useful navigation hint for AI crawlers, but it’s 5% of the job, not 95%. It doesn’t create authority, doesn’t clarify your entity, doesn’t rewrite your content. Anyone selling it as “LLM Optimisation” has sold you a 20-minute job at a 90-day price.
One schema markup pass
Schema helps. One pass doesn’t. Schema needs to match your actual content, be kept current as your services change, and be backed by entity consistency across the web. A one-off JSON-LD injection is a checkbox, not an outcome.
“AI-optimised” keyword stuffing
AI doesn’t work on keyword density. It works on whether your content answers the question directly, whether your entity is unambiguous, and whether other trusted sources corroborate what you say. Stuffing “best AI consultant” into your H2 repeatedly is 2010 SEO wearing a new hat.
“Guaranteed #1 in ChatGPT” promises
Nobody can credibly guarantee a specific output from a probabilistic AI system. Answers vary by query, location, user history and model update. If someone’s guaranteeing rank, they’re either lying or they don’t understand the technology.
Two-week “AI transformations”
A lot of what makes you visible to AI overlaps with what makes you visible to Google — clean structured content, consistent entity data, mentions from trusted sources. That’s 3–6 months of steady work. Anyone selling a two-week fix is setting you up to be disappointed.
How Visual Network does it
A 90-day programme in four phases. Measured, documented, honest.
Diagnose
Weeks 1–3. Baseline audit across ChatGPT, Perplexity, Gemini, Claude. Technical crawl, schema audit, entity audit, competitor benchmark. Documented “before” state with screenshots.
Remediate
Weeks 4–7. Technical and content fixes. Schema, llms.txt, quotable rewrites. FAQ sections structured for AI retrieval. Content lifted out of PDFs into HTML.
Build authority
Weeks 8–11. Review platforms, industry directories, trade press, guest content, podcasts — the sources AI models actually cite. Seeded deliberately, not sprayed.
Re-measure
Week 12. Same queries as week 1, re-run across all four assistants. Documented before/after. You can point at the delta.
What we promise isn’t a guaranteed #1 in ChatGPT. It’s a measurable improvement in AI visibility for an agreed shortlist of queries, against a documented baseline. That’s the honest version.
Who this is for
UK SMEs
Trades, professional services, hospitality, e-commerce. If you’re a local service business whose next customer is currently Googling you, they’ll soon be ChatGPT-ing you. You want to be the one AI recommends.
In-house marketing managers & small agencies
Teams who want AI/LLM visibility expertise bolted onto their existing stack without hiring a full-time specialist. You own the strategy, we own the optimisation.
Less good fit: businesses with no existing web presence, paid-media-only operations, or anyone looking for guaranteed rankings.
Frequently asked questions
Is LLM Optimisation the same as SEO?
Overlapping, not identical. The hygiene fundamentals — structured content, clean entity data, authoritative mentions — help both. What’s distinct is writing pages as directly quotable answers, making your business entity unambiguous to AI, and getting cited in the narrower set of sources AI models actually quote. If your SEO is weak, we fix that too; they’re worked on in parallel.
Will adding llms.txt to my site get me recommended?
On its own, no. /llms.txt is a navigation hint to AI crawlers — useful, but a small part of the job. Being recommended depends on content structure, entity clarity and third-party authority. Anyone selling llms.txt as a standalone solution is charging you for 20 minutes of work.
How long until I see AI starting to recommend me?
Structural fixes (schema, quotable rewrites, llms.txt) can surface in AI answers within weeks. Authority and citation work takes 2–3 months to compound. That’s why the core programme is 90 days, not 14.
Can you guarantee I’ll be the #1 recommendation in ChatGPT?
No. AI outputs are probabilistic and vary by query, location and model update. What we can promise is a measurable improvement for an agreed set of queries against a documented baseline — and if it didn’t move, we say so.
Do I need to pause my Google SEO to do this?
No. Most of the technical and content work improves both simultaneously. We won’t do anything that damages your Google visibility — that would be daft, given Google is still where most of your traffic comes from today.
What if I just want to see if I have a problem before committing?
Run the Free AI Visibility Check. Ten minutes, no sales call. If there’s no problem to fix, we’ll tell you so. Most sites have a problem — but not all, and pretending every business urgently needs our services would be the same dishonesty we criticise on this page.
Start with the free check
Ten minutes. No sales call. No pitch. You’ll see whether AI is currently recommending you — and if not, what’s in the way.
Run my free AI visibility check