Getting cited by ChatGPT and Perplexity: what actually works
« ChatGPT recommends our competitors, not us. » The cause is rarely mysterious: your pages are hard for a machine to read, and your claims cannot be checked anywhere.
Make your pages machine-readable
A model crawling the web receives HTML loaded with menus, scripts and banners. The worse the signal-to-noise ratio, the less of your page survives.
Three moves, in order of effect:
- Serve a Markdown version of every page, declared in the head and available at the same address when an agent asks for it;
- Publish an
llms.txtat the root: the site’s table of contents on one page, one description line per link; - Explicitly allow AI crawlers in
robots.txt— and check that your host does not block them by default, which happens more often than you would think.
Write to be quoted
Generative engines quote self-contained passages. A paragraph that only makes sense after the previous three will never be picked up.
A complete answer at the top
Sixty to ninety words, right after the heading, answering the page’s implicit question on their own. That is the most quoted passage on any page.
One heading = one question
« How much does an AI agent cost? » rather than « Our pricing model ». People ask questions; models look for answers.
Tables
A comparison table is a fact that is already structured. Tables are over-represented in citations — the best effort-to-result ratio on this list.
Be checkable
A model choosing between two sources prefers the one that can be verified. Date your pages, name the author, cite sources with a link, and state what you do not do.
A page that says « market leader » is quotable nowhere. A page that says « three to five days, measured across eleven jobs » is quotable everywhere.
That is why our industry pages show ranges rather than promises, and why they say plainly what an agent will never do.
What is a waste of time
Stuffing pages with keywords. It never worked well for classic search; it does not work at all for generative engines, which summarise meaning.
Publishing near-identical pages. Thirty city pages that differ only by the city name collapse into one — and it will be a competitor’s that gets cited.
Paying to « be in the AIs ». There is no slot to buy inside a generated answer. What exists is readable, dated, checkable content — and patience.
- What we do — including the visibility system
- AI for e-commerce — product pages, readable by machines
- The survey of your ten systems — free, delivered within 72 hours
Other notes
Your document base, or why the agent answers beside the point
An agent is only worth what it can read. The four defects that produce answers beside the point — duplicates, undated versions, scanned PDFs, implicit context — and the preparation that fixes them.
Responsible AI · 14 August 2026« What if it makes things up? » — getting an agent to cite its sources
A model produces a plausible sentence, not a true one. The four designs that make invention visible, and the single rule that matters: every claim points to its source.
Systems · 4 September 2026AI agent or plain automation: the £20,000 question
A third of the use cases brought to us need no model at all. How to tell a workflow from an agent, and why getting it wrong is expensive in both directions.