Only 28% of Americans say they trust AI assistants — yet the people building those assistants have never had more consideration from young buyers. That gap sits underneath almost every story this week, and it's the thing I'd want you thinking about if you sell online. Because right now there's a whole industry springing up selling you "AI search fixes", and a fair chunk of it is nonsense.
Let me walk you through what actually happened this week and what's worth your money.
The Cats.txt Experiment That Exposed a Whole Cottage Industry
If you've had someone email you recently about adding an "llms.txt" file to your site so ChatGPT and friends can understand your products better, this is the piece to read. Someone tested the four arguments used to sell llms.txt by applying them to a text file about cats. All four arguments passed. Which tells you the arguments were never really evidence — they were vibes.
What this means for your shop: you're going to get pitched a lot of AI-visibility tactics over the next year. Some will cost you a few hundred quid, some a few thousand. The test I'd apply is simple — if the same reasoning would justify uploading a file about cats, it isn't a reason. Ask the person selling it: what would prove this doesn't work? If they can't answer, keep your money.
I'm not saying every GEO tactic is rubbish. I'm saying the evidence bar in this space is currently on the floor, and you're the one paying for it.
The Boring Truth About AI Search: Good Work Still Wins
Here's a refreshingly unglamorous take — AI search only feels like a brand new sport to people who were never really training. The fundamentals that made you findable in Google still make you findable in AI answers: clear product information, genuine expertise, structured data that actually describes what you sell, a site that isn't a mess.
What this means for your shop: if your product pages have thin descriptions lifted from the manufacturer, no specifications, no sizing detail, no answers to the questions people actually ask before buying — that was already hurting you. AI search just made it more obvious. The fix isn't a new tactic, it's the work you've been putting off.
Going forward, I'd rather see you spend a week rewriting your top 50 product pages properly than spend that money on a GEO audit. One of those compounds.
Gen Z Now Treats ChatGPT and Claude Like Brands — But Doesn't Trust Them
YouGov data shows consideration for Claude and OpenAI roughly doubling among Gen Z, while overall US trust in AI assistants sits at just 28%. That combination is genuinely interesting. Young buyers are using these tools constantly — and simultaneously not quite believing what they say.
What this means for your shop: people are increasingly researching with AI and then verifying elsewhere. They'll ask ChatGPT for gift ideas or a product comparison, get a shortlist, and then go and check those brands on Google, on Reddit, on your actual website. So the AI mention isn't the sale — it's the introduction. If someone gets pointed at you and then lands on a site with no reviews, vague delivery information and no clear returns policy, that low trust does exactly what you'd expect.
The practical version: your Google Ads brand campaign matters more than it did, not less. That verification search is a real moment, and you want to own it.
Reddit and YouTube Are Doing More Work Than You Think
This one's worth understanding properly. AI systems lean heavily on user-generated content — Reddit threads, YouTube reviews, forum discussions — when deciding what to say about products and brands. Not because that content is more authoritative, but because it's where real people talk about real experiences in language that maps neatly onto how people ask questions.
What this means for your shop: if there's a Reddit thread comparing your category and you're not in it, you're invisible in a place that's feeding AI answers. Same with YouTube — an honest unboxing or comparison video from a small creator can end up doing more for your AI visibility than any amount of on-site optimisation.
I'm not suggesting you go and astroturf Reddit. That backfires spectacularly and the community will spot it in minutes. But encouraging genuine reviews, working with small creators in your niche, and actually participating where your customers hang out — that's the version that works. It's slow. It's also durable, which most of the alternatives aren't.
Citation Share Is the New Impression Share
This is the framing I've found most useful all week. For twenty years, impression share was the number that told you how much of the available auction you were capturing. The argument here is that citation share — how often AI systems mention you when someone asks about your category — is becoming the equivalent metric, with a suggested 90-day sprint for measuring it.
What this means for your shop: you can measure this yourself, cheaply, today. Write down the ten questions a customer asks before buying from you. Ask ChatGPT, Gemini and Perplexity each one. Note who gets mentioned. Do it again next month. That's your baseline, and it costs you an afternoon rather than a retainer.
If you're never mentioned and three competitors always are, that's a real gap worth working on. If you're mentioned regularly, congratulations, stop worrying about it and get back to your Shopping feed.
What I'm Watching
The theme running through all of this is that the fundamentals haven't changed nearly as much as the marketing around them suggests. Good product information, real customer conversations, a site people trust — that's what fed Google and it's what feeds AI answers. There's also a useful point doing the rounds from Google's former chief scientist Jeff Dean, who argues the model itself matters less than the context you give it. Same principle, different layer.
What I'd watch over the next few months is whether any of these AI-visibility tactics develop actual evidence behind them. Some will. Most won't. Until then, the safest place to put your money is the work that pays off either way — better product pages, more genuine reviews, a Shopping feed that's actually accurate, and brand search you properly own.
If you're getting pitched something and you're not sure whether it's real, send it over. I'll tell you straight.