By the time this is published, ChatGPT will have crossed eight hundred million weekly users. For an increasing share of those users, ChatGPT has replaced Google as the first place they ask questions โ including the considered-purchase questions that lead to a shortlist of suppliers. If youโre not in those answers, youโre not on the shortlist.
This is what actually works to get cited โ in 2026, in the UK, for firms with the kind of substance buyers expect.
1. Be a recognised entity, not just a website
ChatGPT cites sources it can verify. That means your firm needs to exist as a clear entity โ named experts with bios, consistent address and identity signals across the web, proper Organisation and Person schema, sameAs links to LinkedIn and professional registers. Entity work isnโt glamorous but itโs table stakes. AI systems wonโt cite a firm they canโt verify.
2. Publish answer-shaped content under named authors
The content patterns that get cited in ChatGPT are remarkably consistent: a clear answer near the top, named claims with verifiable sourcing, plain language, structured headings, FAQs where they fit naturally. Walls of marketing copy donโt get cited. Detailed practical commentary by a named expert does.
The byline matters. โBy The Firmโ signals content marketing. โBy [named solicitor, named adviser, named consultant]โ signals expertise that AI systems weight more heavily.
3. Get the foundational schema right
FAQPage on pages that answer real questions. Service schema on service pages. Person schema on every author. Schema isnโt a magic ranking factor but itโs how AI systems efficiently parse and verify your content. Get it right once and it pays back for years.
4. Optimise for the prompts buyers actually use
Most firms test โwhat is X serviceโ type prompts. The interesting ones are the commercial prompts: โBest [practice area] solicitor in [city]โ, โWho should I use for [specific service]โ, โ[firm name] vs [competitor name]โ. Run them monthly across ChatGPT, Perplexity, Gemini, AI Overviews. Track where you appear, where you donโt, and how the answers describe you.
5. Provide an llms.txt and clean structured exports
Newer AI systems look for an llms.txt file describing your siteโs structure and key content. Clean structured data exports (FAQs, services, locations) help them parse your site efficiently. These are still optional in 2026; theyโll be expected by 2027.
What doesnโt work
Keyword-stuffed content. Generic AI-generated articles published at high volume. Trying to game citation rates with paid mentions. ChatGPTโs training and retrieval systems are explicitly tuned against this kind of content โ publishing it is more likely to damage your visibility than improve it.
If you want to know how your firm actually appears in ChatGPT today, thatโs a thirty-minute conversation. The intro call is here. Or look at how AI search optimisation runs as an engagement.
