GEO / AI Search

The short answer LLM seeding promises something the data does not support. LLM seeding suggests you can plant content in the places models read, then get picked up. One number settles that. Roughly 99 percent of Reddit citations point to individual threads with substantive discussion, rather than to subreddits, profiles or brand-authored posts. So the […]

Is Google sending less traffic than before? On queries where AI features appear, yes, and a randomised experiment published in August 2026 puts the effect at 18.8 percentage points of click-through. Across a whole site the picture is much softer, because most queries do not trigger those features and impressions have grown. The traffic is also not migrating to AI engines, which still account for under 1% of referrals. This article ranks every 2026 study by how well it identifies the effect, rather than by how large its headline is.

How AI search actually works is usually described as a tidy pipeline: you ask, it searches, it reads, it answers, it cites. Each of those five stages behaves differently from the description. In one 2026 pilot the engine searched the web on only 42% of buying prompts. When it did search, it ran an average of 6.3 queries nobody wrote. And in a 2,100-question evaluation across six chatbots, retrieval rather than reasoning drove more than 70% of all errors. This article walks the real pipeline and marks what breaks at each step.

Where ChatGPT and Perplexity get their information about companies is a measurable question, and the answer is that they barely read the same internet. Across 27 million citations analysed in January 2026, Perplexity drew 19.4% of its citations from social sources against ChatGPT’s 5%, while ChatGPT took 15% from institutional sources against Perplexity’s 5%. They also search differently: 91% of ChatGPT’s retrieval queries were unique, against 14% of Perplexity’s. One rewrites your question beyond recognition, the other searches it almost word for word.

You can check if your brand appears in AI search in about twenty minutes, on two engines, without paying for anything. The method is simple and the result is noisier than most people admit. In one 2026 pilot, ChatGPT searched the web on only 42% of buying prompts, so on the rest there was no retrieval event at all. In another, rewording a question changed the answer more than 23% of the time. This article gives you the check, then gives you the four reasons it can lie, so you can tell a real absence from a bad sample.

The research pulls in two directions. Fan-out rewards covering many adjacent questions; absorption research finds longer pages get used. The resolution is that the passage competes, not the page.

Seven documented changes, not seven predictions. Google switched two rich results off, shipped two new AI reports, and added a channel to Analytics. Here is what actually moved and what it means.

I trust signals split in two. An engine can trust your page enough to use it, and still not trust your name enough to say it. Across 3,981 measured appearances, only 13.2 percent were both cited and named.

You can check if your brand appears in AI search in about twenty minutes, on two engines, without paying for anything. The method is simple and the result is noisier than most people admit. In one 2026 pilot, ChatGPT searched the web on only 42% of buying prompts, so on the rest there was no retrieval event at all. In another, rewording a question changed the answer more than 23% of the time. This article gives you the check, then gives you the four reasons it can lie, so you can tell a real absence from a bad sample.