To Get Mentioned by ChatGPT, Stop Fixing Your Website

Imagine applying for a job where the employer reads exactly one reference. You don't get to write it. You don't get to see who does.

That is roughly the position you're in when you try to get mentioned by ChatGPT.

When a customer asks an AI which business to use, most of what it draws on was published by someone other than you. Across three editions of Muck Rack's citation study, earned sources have held between 82% and 89% of all AI citations.

Your own website still matters. It is the smaller half of the job, and almost every guide on this subject is written about the smaller half.

We already covered that smaller half. Getting your own house in order is the prerequisite for everything below: matching business details everywhere, a finished Google Business Profile, answer-first writing, an FAQ section, a real byline.

Do that first. This post is about everywhere else.

What is an AI engine actually reading when it recommends a business?

Mostly things you didn't write. When an AI names a business it assembles the answer from third-party pages: reviews, directories, articles, forum threads, video descriptions, other companies' sites.

Your own pages are one voice in that pile, and not the loudest.

Muck Rack's May 2026 analysis of more than 25 million links from ChatGPT, Claude and Gemini put earned media at 84% of all citations, with paid and advertorial content at 0.3%. Journalism alone accounted for 27%.

A separate Fullintel study with the University of Connecticut, published in February 2026, put journalism higher still, at 47% of citations.

Worth knowing that Muck Rack sells PR software and Fullintel sells media monitoring, so neither is a disinterested party. What makes the direction hard to dismiss is that the earned-media share has barely moved across a year of measurement.

Here's the rough shape of it:

Source type	How much AI leans on it	Can you control it?
Earned coverage on other people's sites	84% of citations, 82% to 89% across three editions	Influence only
Your own website and profiles	The minority share	Fully
Backlinks	0.218 correlation with AI Overview visibility	Partly
Wire press releases	Almost only in industry-trend answers	Fully, and not where your customers ask

Read the row for your own website twice. It's the only thing on the list you control completely, and it's the thing engines lean on least.

Why does being mentioned beat being linked to?

Because engines are counting how often the world talks about you, not how many sites point at you. Across 75,000 brands in the Ahrefs dataset, branded web mentions correlate with AI Overview visibility at 0.664, the strongest of every factor they measured. Backlinks sit at 0.218.

Correlation isn't proof, and nobody outside these companies can see the ranking systems directly. The practical translation is clear enough: a paragraph describing your business on someone else's page beats a link buried in their footer.

Mentions without links count. That single fact changes what's worth chasing, because getting someone to write a sentence about you is far easier than getting them to link to you.

Does ranking on Google still get me cited?

Less than it used to, and this is the number that should reset how you think about the whole problem. Only 38% of AI Overview citations now come from Google's top ten results. A year earlier that figure was 76%.

Ranking first on a query gives you roughly a one-in-three chance of being cited in the answer above it.

Sit with that. Half the connection between "ranks well" and "gets named" has come apart in about a year.

The old plan was to climb the list and let everything follow. The list is no longer the thing standing between you and the customer.

The WhatsApp problem

Most recommendations in Saudi Arabia happen inside private messages, and no AI model can read a single one of them. This is the part that makes the country genuinely different, and the reason a well-loved business can be invisible.

Someone asks the family group which workshop is honest. Three people answer. A name gets forwarded to a neighbour, who forwards it to a colleague. That chain is the most powerful marketing channel a local business has.

So you can be the most recommended barber in your district and still be a blank space to ChatGPT. Your reputation is real. It just isn't written down anywhere a machine can reach.

Nobody should stop that word of mouth. The point is that it doesn't convert itself into machine-readable presence, and it never will.

Somebody has to carry it across deliberately, usually by asking the people who already recommend you in private to say the same thing once in public.

What does "earned media" mean when nobody is writing about my shop?

It means something different here than it does in the studies. A café in Khobar is not getting a feature in a technology publication. Neither is your clinic, your workshop, or your three branches of anything.

When American research says "earned media" it is usually picturing journalism, and that picture doesn't survive the trip.

So rebuild the idea from what actually exists around a Saudi business:

• Google Maps reviews and the questions section. The single richest source most local businesses have, and the one they most often ignore. Reviews that describe what you actually do give an engine something to quote.
• Local and sector directories. Thin pages, but they're indexed, consistent, and free.
• Chamber of commerce and trade association listings. Dull, credible, rarely claimed.
• Supplier and partner pages. If you're a stockist, a certified installer, an authorised service centre, ask to be listed by name. This is the most overlooked one on the list.
• Local and regional press. Sabq, Okaz, and regional outlets do cover openings, expansions, and community events. Smaller than a national feature, and real.
• Video. A supplier's walkthrough, a customer's review, a local creator's visit.
• Forums and public threads. Reddit and X discussions get read by engines constantly.

None of that requires a budget. It requires somebody spending an afternoon asking to be named in places that already exist.

Why is your Arabic footprint different from your English one?

Because engines answer from different sources depending on which language the question arrives in. Ask about a Riyadh clinic in English and you get one set of references. Ask the same thing in Arabic and the engine reaches for different pages entirely.

Most businesses here have never checked both. The usual result is lopsided: a decent English presence assembled for search engines years ago and almost nothing in Arabic, or a strong Arabic word-of-mouth reputation with no written trace at all.

Either way, half your customers are asking in the language where you don't exist.

The fix starts with fifteen minutes and no tools. Ask ChatGPT, Gemini, and Perplexity for a recommendation in your category and city, in English. Ask again in Arabic.

Write down whether you get named, and what the engine seems to be reading when it names someone else. Most owners find the two languages disagree, and the gap tells you exactly which side needs work.

What to do this month

Five things, all free, in order:

1. Run the audit in both languages. Fifteen minutes. It sets your priorities better than any advice here.
2. Claim what already exists. Google Business Profile, directory listings, association pages. Complete them and make the details identical everywhere.
3. Get named on three third-party pages. Pick ones you can realistically reach: a supplier, a partner, a local outlet, an association. Ask to be described, not just linked.
4. Ask for reviews with specifics in them. "Great service" is worthless to an engine. "Fixed my espresso machine same day in Al Khobar" is a quotable sentence about what you do and where.
5. Put one real fact in your own content. A number, a price range, a timeline. Give other people something concrete to repeat, because repetition is the mechanism.

What not to bother with

Wire press releases, at least in your situation. Their citations grew fivefold between July and December 2025, so they are being picked up more, not less. The catch is where they get picked up: almost entirely in industry-trend answers, at 3.5 times the rate they appear in the best-of queries where somebody asks which business to use. Those best-of answers are the only ones a local shop is competing in.

Also skip any tool promising to get you mentioned. AI visibility monitoring is a real product category now, aimed at marketing teams with budgets and dashboards to fill. There's nothing in it a shop owner needs that fifteen minutes of asking questions won't tell you.

How will you know it worked?

Re-run the same audit in three months, with the same queries in both languages. That's the whole measurement system. There's no dashboard, no score, and no way to buy your way up.

Expect it to move slowly. Engines rebuild their picture of you as their sources change, so a listing you fix today shows up in answers weeks or months later.

The businesses that win at this are the ones that did the boring work early and kept their details consistent while everyone else waited for a shortcut.

One thing makes the boring work less boring. The details an engine finds about you have to match across every place they appear, and that is far easier when your business information lives in one system instead of scattered across six.

Swarlet keeps that record in one place, which turns "update everywhere" into something you can actually finish. Join the waitlist to get it when it opens, and while you wait, go get your own site in order.

Figures in this post are current as of August 2026. AI citation research is moving fast.