THEY SAID*

GETTING FOUND / THE ANSWER / 22 JUL 2026 / 4 MIN READ

How do you get your agency cited in ChatGPT?

You get cited in ChatGPT by appearing in the trade-press articles and category anchor pages the AI engines already trust - not by publishing more content on your own site.

The short answer: appear as the named expert inside the specific trade-press articles, directory pages, and category roundups that ChatGPT already treats as authoritative for your category - not by publishing more content on your own site.

That's the mechanic. The rest of this post is the reasoning, and the workflow.

Why your own website is the wrong lever

The instinct is to fix your own website. Write more blog posts, add more service pages, cram the phrase 'creative agency London' into the H1. That's SEO logic applied to a system that isn't SEO.

ChatGPT, Perplexity, Claude, and Google's AI Overviews don't rank websites the way Google ranks websites. They synthesise answers from a small number of sources they've decided are reliable for a given category, and they cite from that shortlist. For UK agency categories, that shortlist is not your website. It's Campaign, PRWeek, The Drum, LBB, Creative Boom, and a handful of aggregator directories.

If you're not in those sources, more content on your own site doesn't move the answer. If you are in those sources, the answer changes.

Where citations actually come from

Three surface types drive most AI citations in agency categories:

One - trade-press bylines and expert comment. When an AI engine is asked 'what's the best independent PR agency for fintech in London?', it draws on articles that name agencies in that category. Not press releases. Editorial pieces where a named journalist made a judgement. That means Campaign's agency-of-the-year features, PRWeek's sector roundups, The Drum's world creative rankings, Creative Boom's Booms & Shakes column, and any list-format article that a publication publishes twice a year.

Two - category anchor pages. These are the pieces of content the engines rely on when the query is a category definition rather than a name. 'What is generative engine optimisation?' gets answered from PRmoment, Forbes agency-council posts, and a handful of others. The engines return the same short list of URLs each time. Appearing in that list, as an expert quoted or a source cited, is what gets your agency named the next time someone asks a related question.

Three - directory and aggregator pages. Clutch, Sortlist, The Drum Recommends, Alliance of Independent Agencies member listings, PRWeek's Agency Business Report tables. These matter less than trade press but more than most agencies think. AI engines use them as sanity-check sources - the 'is this agency real?' pass before they name you in an answer.

The workflow that moves the needle

The mechanic is boring on purpose. It is four steps in sequence:

  1. Audit what the engines currently say. Run twenty prompts across ChatGPT, Perplexity, Claude, and Google's AI Overviews, using the queries a prospective client would actually type. Log the sources cited. This is where you learn which trade titles are driving your category and where the gap in your presence is.
  2. Place expert comment where the gap is loudest. If Campaign is showing up in the citations for your category but you're not quoted in Campaign, that's the first placement to earn. Expert comment lands faster than bylines and is what the engines index first.
  3. Get onto the list articles. Every UK trade title runs a predictable rota of list features - agencies to watch, best independent PR agencies, agency-of-the-year shortlists, sector roundups. Editors write these pieces in a two-week window and are looking for named contenders. Being useful when they ask, and being the agency that pitched itself for the piece two months earlier, is what gets you onto the list.
  4. Re-run the audit at eight and twelve weeks. The engines re-index on a delayed cycle. You will not see the shift the day the article publishes. You will see it four to eight weeks later, when the engine's next indexing pass has caught up with the new URLs.

What doesn't work

Publishing more posts on your own agency blog. Adding structured data to your own about page. Buying followers. Running paid amplification on your own content. Commenting on LinkedIn posts about GEO.

Every one of those is a lever inside a system where the levers are outside. It feels productive. It changes nothing about what the AI engines return.

You are not optimising your website. You are optimising the sources the AI engines quote when they answer questions about your category.

The honest timeline

First movement in the audit takes eight to twelve weeks, sometimes longer. Sustained presence takes two to three quarters. There is no quick version. Anyone offering you a two-week AI-visibility programme is selling on-page SEO with a different label on the invoice.

The good news is that once you're in the citations, you're sticky. The engines pass along their trust to the sources they've already learned to trust. That's the compounding curve that makes this worth doing properly the first time.

If you want the diagnostic that sits underneath any sensible programme, we've written the entity audit spec separately - it's the piece of work that has to happen before any of the above is worth commissioning.

WRITTEN BY

Fayola Douglas, founder of They Said

MORE IN GETTING FOUND

Want this engine running on your agency?

Book a call ↗