Skip to content
Matrex Analytica

Map the authority signals before you map the AI stack

The instinct, when a new class of tools arrives, is to plug them in first and work out the strategy later. With AI visibility this gets the order wrong. Before adding another layer to your brand stack, decide which authority signals you want AI systems to repeat about you. If you do not define them, the systems will assemble their own version from whatever third-party material happens to be dense enough to notice.

Semrush published its 2026 AI Visibility Index UK and US benchmark on 13 October 2026, covering seven months of fixed monthly sets of 2.5 million UK and 25 million US prompts across ChatGPT, Gemini, Google AI Mode and Google AI Overviews. The headline finding for anyone running a UK brand is that visibility is fragmented across the platforms rather than consolidated. Any two of the four AI systems shared only 63 to 70 of their top 100 mentioned brands on average in the UK, against 70 to 78 in the US. Only 24 UK domains sat in the top 100 mentions on all four platforms every month, compared with 36 in the US, and just 15 of those were shared across both markets.

The deeper number is the gap between being mentioned and being cited. Across 22 categories and four platforms, fewer than half of the top 100 mentioned brands also appeared among the top 100 cited domains. In the UK, that overlap ranged from 36.7 to 45.3 domains out of 100 by month. Mention and citation are not the same economic asset. A mention places you in an answer; a citation sends the user to your material and lets you shape what they read next. Semrush points to the Tesco and Asda contrast in July: Tesco was ChatGPT's 25th most-cited UK domain while Asda ranked 606th, though on Gemini the two sat much closer, at 29th and 60th. Two direct competitors, one category, four systems, four different pictures.

Why this reframes the brand problem

A brand's presence used to be something you could audit in roughly one place at a time: your own site, a search results page, a press clipping file. AI answers break that habit. Each system assembles a response from a different mix of sources, and the mix shifts by platform, market and category. News publishers carry particular weight in the UK: on ChatGPT, news sites averaged 29.5 per cent of citations among the 100 most-cited UK domains against 19.6 per cent in the US, and on Gemini news led UK categories at 16.6 to 23.0 per cent monthly. If the facts a news site chose to repeat about you two years ago are the densest material available, those are the facts the model will tend to echo now.

That is why the order of operations matters. The tools in an AI brand stack, monitoring platforms, content generators, schema editors, prompt testers, all optimise against some target. If the target is implicit, the stack will push you toward whatever is easiest to produce more of, usually volume on your own channels. The Semrush data suggests that volume is not the binding constraint. Overlap between mention and citation is. Which specific claims, numbers, definitions and attributions do you want surfaced when a UK buyer asks a model about your category? That list is the brief every other tool should serve.

A practical sequence

A short written brief is enough to start. Three parts tend to matter.

  • The five or six authority signals you want repeated: a defining capability, a measurable outcome stated the same way each time, the names of the people who speak for the work, the categories you want to be placed in, and the comparisons you are willing to be drawn into.
  • The sources outside your control that currently carry those signals, or carry rival ones. For UK brands this almost always includes a handful of national and trade news titles, given their citation share.
  • The gaps between the two: signals you want that no third-party source states clearly, and signals third parties state about you that you would not choose.

Once that brief exists, the tooling decisions become narrower. Monitoring is set to track those specific signals across the four systems rather than a generic share of voice. Publishing effort is weighted toward the gaps, not toward the channels that are already dense. Relationships with the news and trade sites that disproportionately feed UK answers are treated as part of the brand asset, not as a separate PR line item.

The commercial stake

Semrush cites Adobe data in the same report stating that AI-referred visitors convert 60 per cent better than non-AI traffic and generate 53 per cent more revenue per visit. Those figures come from Adobe through Semrush rather than independent verification, and the study measures mentions and citations rather than causal attribution, so treat them as directional. The direction is still useful: the traffic that AI systems do send is more valuable per visit than the traffic they replace, which raises the cost of being mentioned without being cited, or cited inconsistently across platforms.

There are honest limits to what one index can tell a brand owner. The UK dataset is a tenth the size of the US one and some findings will be noisier. The observation window ends in July 2026 and platform behaviour has probably moved since. Semrush sells AI visibility tooling, which colours how it frames integrated strategies. None of that changes the structural point the data makes clearly enough: UK AI visibility is more fragmented than US visibility, mentions and citations diverge, and the sources that feed the answers are not the ones most brands have been optimising for.

The useful move before buying another tool is quiet and unglamorous: write down the signals you want AI to repeat, in the exact words you want it to use, and check which third parties already carry them. Everything in the stack can then be judged against that list.

Source: Semrush, 2026 AI Visibility Index UK and US benchmark, 13 October 2026.

Source: https://www.semrush.com/news/474244-semrush-releases-2026-ai-visibility-index-revealing-its-first-uk-benchmark