Ask a comms director in Dubai what their last campaign achieved and you’ll get a familiar answer: placements secured, reach delivered, a logo bar refreshed for the website footer. Now ask ChatGPT what their brand does, who founded it, and whether it’s any good.

That second answer is the one their next customer will actually see. And most brands in this region have no idea what it says.

PR has spent two decades optimising for a reader who is quietly disappearing. The buyer journey that used to run through Google results and news homepages now increasingly runs through an AI answer: a few confident sentences from ChatGPT, Gemini or an AI-powered search result, synthesised from whatever sources the model considers trustworthy.

Pew Research Center found that when an AI summary appears in Google results, users click a traditional link just 8 per cent of the time, half the rate of searches without one, and click the sources cited inside the summary in only 1 per cent of visits. Your prospect doesn’t scroll past ten blue links anymore. They get one paragraph. Either you’re in it, accurately, or you don’t exist.

This changes what coverage is for. Until recently, a placement was something a human read that week, and that was the end of it. Now it’s also training data and citation material: a permanent vote on how machines will describe you for years. The press release you distributed in 2023 is still shaping what ChatGPT says about your brand in 2026. Coverage has stopped being a moment and become infrastructure.

The uncomfortable part for our industry is that the metrics we sell don’t measure any of this. “Featured on 400 sites” was always a soft number. In the AI era it’s actively misleading, because language models don’t count placements – they weigh them. A model deciding what’s true about your company treats a report in renowned media outlets as strong evidence. By every indication it treats the four-hundredth syndicated copy on a site no human has ever visited as noise, if it ingests it at all.

PR people have always known this instinctively. It’s why we tier media lists – five points for a Tier 1 hit, three for the trades, one for the long tail. What’s changed is that the machines have adopted our scoring system and made it ruthless. A handful of placements in outlets the models treat as authoritative now does more work than hundreds of syndicated copies.

The regional implications are bigger than most Gulf brands realise. Roughly 44 per cent of Common Crawl, the web corpus underpinning most major AI models, is English-language content, drawn heavily from Western sources, which means Middle East companies are already under-represented in the material that shapes AI answers. A regional champion with a thin footprint in trusted media is ceding its own narrative to whatever fragments the model can find. Ask ChatGPT about your category in the UAE and see who it names. If it isn’t you, that’s not a branding problem. It’s a data problem.

Start with the KPI. Stop reporting placement counts and start reporting weighted authority – and, increasingly, the answer itself. The most honest PR dashboard in 2026 includes a screenshot of what the major AI assistants say about your brand this quarter versus last.

The strategy follows from that. Fewer, better placements beat broader, thinner ones. A data-led story that earns genuine pickup in two or three trusted regional and international titles will do more for your machine-readable reputation than any volume package – and I say that as someone who sells distribution for a living.

Then there’s the housekeeping. Accuracy has become a compounding asset. A wrong founding date or a stale executive name in a widely syndicated release doesn’t just sit in an archive anymore; it gets repeated back to your customers with total confidence by a machine that doesn’t know it’s wrong. Correcting the record in authoritative sources is now cheaper than living with the error.

Keep the ‘as featured in’ logo bar if you like. It just answers a question nobody asks anymore. The question your market is asking is being answered by a machine, in one paragraph, right now.

You should probably find out what it says.

By Sam Allcock, Founder of Digital24