There's something shifting in how customers find products this week, and it's happening well before anyone types a search into Google. A study of how ChatGPT actually behaves when it recommends products has landed, and it changes the shape of the problem for anyone selling online. Alongside that, Google Ads has quietly announced it's pulling a targeting setting many accounts still rely on. Here's what's happening and what it means for your store.
ChatGPT Decides Who to Recommend Before It Even Searches
This is the one worth reading twice. The research found that when someone asks ChatGPT for a product recommendation, ChatGPT often names specific brands in the search query it writes for itself — before it has fetched a single page. It's not searching neutrally and then deciding. It has already decided, and the search is largely confirmation.
Being named in that internal query was found to be worth roughly 33 times more than simply appearing in the results it later reads.
What does that actually mean for your business? It means the old model — rank well, get found, get picked — doesn't fully apply here. If ChatGPT's underlying model doesn't already associate your brand with your product category, you're not in the running before the race starts. You're competing on how well-known you are in the training data and in the wider web conversation, not just on how well your product page is optimised.
Practically, this pushes weight back onto the unglamorous stuff: getting mentioned on comparison sites, review roundups, forums, trade press, anywhere your category gets discussed. Those mentions are what build the association. It's a slower game than bidding on a keyword, but it's increasingly the game that decides whether you're considered at all.
Google Is Removing Language Targeting From Search Campaigns
From September 2026, Google is changing how language targeting works in Search and Performance Max campaigns. The setting that lets you restrict who sees your ads based on their browser and interface language is going away.
If you sell only in the UK to English-speaking customers, this probably won't shake your world. But if you're running multi-language campaigns, or you've been using language targeting as a crude filter to keep out traffic you don't want, this matters. Google's argument is that its systems can now work out relevance from the query itself rather than needing you to declare it. My view: that's true more often than it used to be, but it removes another lever from your hands.
The practical step is to check your campaigns now rather than in September. If language targeting is doing real work in your account — keeping irrelevant international clicks out of a UK-only budget — you need a replacement plan. Location targeting, negative keywords and tighter ad copy all have to carry more of that load once the setting disappears.
OpenAI Says robots.txt Might Not Stop Its Fetching Bot
New data shows ChatGPT's page-fetching bot is reaching sites that have explicitly told it not to. OpenAI's own documentation explains why: when a user asks ChatGPT to look at a specific page, OpenAI treats that as a user-initiated request rather than automated crawling — and robots.txt is aimed at crawlers.
Whether you think that's reasonable depends on where you sit. For most e-commerce businesses, though, this is arguably good news rather than bad. If a customer asks ChatGPT about your product and it can actually fetch your page, it sees your real prices, your real stock levels, your real delivery terms. Blocking it means ChatGPT falls back on whatever it thinks it remembers about you, which is often out of date.
The thing to check is whether your product pages actually make sense to a bot reading them. If your prices, delivery information and stock status only appear after JavaScript loads, a fetching bot may see very little. That's worth testing.
Google Analytics Adds Campaign Benchmarking
Google Analytics has added campaign benchmarking, letting you compare your performance against aggregated data from similar businesses. Alongside that, there's new information about the size of ChatGPT's search index, and Google has refiled its legal complaint against SerpApi with revised terms.
Benchmarking is the one that affects day-to-day decisions. Knowing your conversion rate is 2.1% is useful. Knowing whether 2.1% is strong or weak for businesses like yours is considerably more useful — it's the difference between "we should fix the site" and "the site's fine, we need more traffic."
The caution I'd add is that benchmark data is only as good as the comparison group. "Similar businesses" is doing a lot of work in that sentence. Treat it as a rough steer, not a verdict on your account.
Value Inflation: When Your Conversion Numbers Are Lying
This piece covers auditing "value inflation" — where the conversion values feeding your Google Ads bidding don't reflect actual revenue. Duplicate conversion actions counting the same sale twice, values recorded before discounts, refunds never subtracted, micro-conversions given a made-up value that then gets treated as real money.
Why does this matter so much? Because Smart Bidding does exactly what you tell it. If your account thinks a sale is worth £120 when it's actually worth £70 after discounts and returns, the algorithm cheerfully overspends chasing it. You get a beautiful ROAS figure in the interface and a disappointing bank balance.
If you've inherited an account, or changed your checkout, or added a new conversion action in the last year, this is worth a proper look. Verify the numbers before you touch the bid strategy — otherwise you're optimising towards fiction.
What I'm Watching
Two threads running through this week. The first is that AI systems are increasingly deciding who gets recommended based on brand association built up over years, not on what you did to your product page last Tuesday. The second is that Google keeps removing manual controls on the assumption its automation handles it better.
Both point the same way: less control over the mechanics, more weight on the fundamentals. Clean data, accurate conversion values, a brand people actually talk about. The levers are disappearing, but the things that were always true are getting more important, not less.
If you're not sure whether your conversion values reflect real revenue, that's the single most valuable thing to check this month. Everything downstream depends on it.