A screenshot is not a standing
By Gail Macleod ·
One AI answer about your brand is a coin toss, not a position. Measure how often you are named, by which engines, in which markets, over weeks.
The signal
On 13 October, Pineprompt published a 14-day benchmark of seven AI search engines: 200 non-branded buyer prompts, run daily from 30 September to 13 October, 18,792 answers in all. The brands named for the same prompt changed from one day to the next by 36% on ChatGPT, 45.1% on Google AI Overviews, 55.8% on Gemini and 68.8% on Perplexity. Of the 8,269 brands seen, only 12.9% were named by every engine, and 46.2% turned up on just one. (Pineprompt — AI Search Visibility Benchmark 2026)
The same day, Semrush's 2026 AI Visibility Index added its first UK benchmark and found that the four big AI platforms agreed less in the UK than in the US on which brands to mention and which sources to cite. Only 15 domains sat in the always-visible core of both markets. Its verdict: there is no single, global way to be visible to AI. (Semrush — 2026 AI Visibility Index, first UK benchmark)
Search Engine Land had already collected the wider evidence. Ahrefs saw AI Overview content change about 70% of the time between checks, and a University of St. Gallen study found about 65% of cited sources changed day to day. The headline: a snapshot isn't a trend. (Search Engine Land — Your AI search visibility keeps changing)
The so-what
Someone on the team asks ChatGPT about you, you come up, and the screenshot goes round the group chat. Or you don't come up, and someone wants budget moved by Monday. So What Law: one answer is a coin toss, not a position. Fluency is not authority here either: the answer reads confident whether you were in it by chance or by design. If you sell in more than one country, a result from one engine in one market says very little about the next one.
The distinction
A screenshot is a moment. A standing is a pattern. A moment tells you what one engine said once. A pattern tells you how often you are named, by which engines, in which markets, for which buyer questions, and whether that holds over weeks. Confusing the two produces both false alarms and false comfort. It also hides the part that does stay put: the facts the engines keep returning to. You can't control which brands an engine names on a given Tuesday. You can control whether the facts it finds about you are verified, current and the same everywhere you publish.
The move
This week, swap the screenshot for a small log. Write (1) ten questions a buyer actually asks before hiring someone like you, (2) the engines they use and the markets you sell in, and (3) a fixed day each week to ask every question again and note whether you were named and which page was cited. Keep a line for what you published or corrected that week, so you can tell your own work apart from the platform shuffling. After four weeks, report a rate with a range, not a single score. Say it out loud: "A screenshot is not a standing."
Quick answers
Why does my brand appear in AI answers one day and vanish the next? AI answers are non-deterministic and engines change their sources without notice. Pineprompt measured day-to-day brand change of 36% to 68.8% depending on the engine, for the same prompt.
How should I measure AI visibility? Use a fixed set of buyer questions, check each engine separately, repeat on a schedule, and report how often you are named over several weeks rather than one result.
Does visibility in one country carry over to another? Not reliably. Semrush found AI platforms agree less in the UK than in the US on which brands to mention, so each market needs its own check.
If one screenshot is currently your AI strategy, AI Visibility Clean-up puts one verified set of facts behind every answer and gives you a before-and-after report. If you'd rather talk it through first, book a free 30-minute call.