Semrush turns AI visibility into a score—but debate flares

By late 2026, the measurement problem had become visible in both studies and product design: according to Analytics Insight, Semrush’s 2026 AI Visibility Index examined more than 126 million real U.S. prompts across four major platforms, while its toolkit was built on more than 317 million AI prompts and responses across 117 regional databases with daily updates. That scale mattered because the index showed cross-platform variability in whether brands were named at all, and Semrush responded with a distinct AI Visibility Score, defined as measuring how consistently and prominently brands appear across relevant AI answers on a 0-to-100 scale using mentions, citations, sentiment, and share of voice.

The market’s response was rapid productization, not just theory: Semrush’s August agency guide explicitly listed the “7 best AI visibility tracking tools for agencies” and described Semrush’s AI Visibility Toolkit and Enterprise AIO as agency platforms to “monitor and improve their clients’ prominence,” while positioning the former as tracking up to 25 prompts by default and the latter as monitoring visibility across 11 AI search platforms from over 289 million prompts. In practice, Semrush said a domain entered into Visibility Overview returned mentions, citations, and cited pages in under three seconds, while share-of-voice reporting could show Pet Libro at 14.3% against rivals with stronger sentiment, underscoring why rankings alone no longer captured brand presence.

Brands need a baseline because AI visibility is not one market but several overlapping systems that answer the same buyer question differently. Semrush says its data comes from a database of more than 289 million prompts and responses across ChatGPT, Gemini, Google AI Overviews, and AI Mode, covering 40+ regional databases, and reports that each platform assembles its answers from a different mix of sources. In its 2026 AI Visibility Index, analyzing 126 million U.S. AI search prompts collected between January and April 2026, only 36 of 1,200-plus brands consistently ranked across all four engines. Spiralytics adds that 5% of the citations in its Philippine answer set pointed to websites owned by the brands being studied.

That fragmentation makes one-off checks useless and pushes brands toward repeatable, multi-engine auditing with composite metrics. Semrush’s workflow says to monitor AI visibility fluctuations daily and uses AI Visibility, a benchmark score from 0 to 100 that measures how prominently and reliably a brand is featured in relevant AI-generated answers, while LLM Scout found ChatGPT, Claude, Gemini, and Perplexity disagreed on whether to name a tracked brand in 27.2% of cases across 749 identical queries. Daryo89’s side-by-side scans and self-test reflect the same need, with its own reports showing 100 GEO, 100 SEO, and 100 Lighthouse.

The behavioral shift is already large enough to reorder marketing priorities because discovery is moving upstream into AI answers and recommendation systems before a brand ever gets a click. HackerNoon captured the break plainly: “In 2026, a growing share of buyers never see a ranked list at all,” while another analysis argued “we no longer search for the content we consume,” noting that 82.4% of viewers arrived via YouTube recommendations and only 5.2% via search—evidence that discovery is increasingly mediated by algorithms, not classic query-and-click journeys. The platform response underscores the scale of that transition: in Alphabet's Q1 earnings call, Google announced “a 43% increase in AI spending to $17 billion” and “committed a total investment of $75 billion over 2020 25.”

That makes site visits and conventional search metrics weaker as primary indicators, because AI is now shaping both research behavior and downstream outcomes before traffic appears in analytics. In a survey of 5,000 U.S. consumers, 55% used AI for research and 39% for online shopping, while traffic from generative AI sources rose 1,200%; those visitors also spent 8% more time, viewed 12% more pages, and bounced 23% less than traditional search users. Razorfish likewise found 94% of U.S. AI users would consider an unfamiliar brand if an assistant recommended it, showing recommendation visibility now matters as much as visits.

AI visibility scores are disputed first because the measurement layer itself is unsettled. B&T noted that the IAB had only just “released the first standardised framework for AI visibility measurement” even as “more than 20 companies hawking AI visibility tools” were using different methods and producing different answers for the same brand; that helps explain why many practitioners now describe these outputs as directional, not decision-grade. On November 3, 2026, Darshan Modi of Mavlers Agency similarly argued agencies should avoid promising “guaranteed mentions, citations, recommendations or leads within a fixed timeline,” because “Those outcomes depend on platforms the agency” does not control.

The second dispute is over what these scores actually predict. University of Toronto researchers found the same queries produced overlap with Google’s top 10 results ranging from just 4 percent on GPT-4o to 15.2 percent on Perplexity, while a 167,551-citation study found only 14.3 percent of AI citations went to brands’ own sites; with most references landing on third-party sources, a visibility gain does not cleanly map to traffic or conversion. That is why marketers remain split: in a Scrunch and Scribewise survey of 602 U.S. professionals, 62 percent said leadership still judges them on SEO or web traffic, even though “Healthy search rankings can hide an AI search problem.”