How to get your business recommended by ChatGPT

Make your website readable by a machine, get mentioned on the pages assistants read about your category, and publish a page for each question your customers ask. There is no switch to flip and no one can guarantee placement — anyone who does is guessing — but those three move the odds, and the first can be done this week.

Step 1 — Make your site readable (this week)

This is the part entirely in your control, and in our audit of 82 UK businesses it was where almost everyone was losing points.

1. Add structured data. schema.org markup for your business type, address, prices and services. It was the single biggest gap in every sector we measured (0.25–0.32 out of 1). It tells a machine plainly what you are instead of leaving it to infer.
2. Publish your prices as text. Not in a PDF, not behind a form. 57% of the wedding venues we audited showed an assistant no price at all — so when someone asks for venues under £8,000, those venues simply cannot be in the answer.
3. Make sure your content loads without JavaScript. Turn it off and look at your site. Whatever disappears may be invisible to an assistant.
4. Allow the AI crawlers. Check your robots.txt does not block GPTBot, ClaudeBot, PerplexityBot or Google-Extended, and that your firewall lets them through.
5. Publish an llms.txt. A short plain-text file at your site root summarising what you do. Only around 15% of businesses we audited had one. It takes ten minutes.

Step 2 — Get onto the pages assistants read (this month)

When an assistant searches the web, it reads a handful of pages and largely names who those pages name. So the question is which pages in your category it reads, and whether you are on them.

• Directories and listings. Google Business Profile, Bing Places, Apple Business Connect and the main directory for your sector. Complete, consistent, with the same name, address and phone number everywhere.
• Guides and "best of" lists. Find the articles in your category that assistants cite and ask to be included. In our research the cited guides were often from small specialist sites, not big brands — very approachable.
• Reviews. Recent, specific reviews on the platforms your customers use. They are read, and they carry the detail assistants quote.

Step 3 — Answer the questions (ongoing)

The pages assistants cite are often titled with the question itself. Find the questions your customers ask — your enquiry inbox is the best source — and publish a genuine answer to each, on your own site, with the question as the heading. Specific beats general: a page on "dog-friendly wedding venues near Cheltenham" competes for a question a general "our venue" page never will.

What does not work

• Asking the assistant to remember you. Conversations do not change what it tells other people.
• Stuffing pages with keywords. Assistants summarise meaning; repetition does not help and reads badly to the humans who arrive.
• Anyone guaranteeing placement. Nobody controls these answers. What can be measured is whether your changes moved them.

Measure before and after

Run a check before you change anything, then again a few weeks later. Without a baseline you cannot tell whether your work did anything — and because answers move on their own, you need more than one data point to tell a real change from noise. Ask All Four does both: a free check now, and weekly monitoring that emails you when an assistant changes its mind.

Questions people also ask

Can I pay ChatGPT to recommend my business?

No. There is no paid placement in the answers of ChatGPT, Perplexity, Gemini or Claude for this kind of question.

How long until ChatGPT recommends me?

It varies by assistant. Perplexity searches the web live and can reflect new pages within weeks of them being indexed. Answers drawn more from training data change when the model is updated.

Do I need an agency to do this?

Not for step one — it is a few hours of work on your website. Steps two and three are ongoing, and are where specialist help can save time.

What is llms.txt?

A plain-text file at the root of your site summarising what you do, for language models to read. It is a convention rather than a standard, but it is free and quick.