ChatGPT handled 70% of US AI prompt volume in January 2026. By June, that had fallen to 50%, according to Comscore's Q2 2026 AI Intelligence Report. Gemini's share nearly doubled over the same six months, from 17% to 30%, and Claude's grew more than fivefold, from 2% to 11%. If your AI Share of Voice tracking still treats ChatGPT as the only engine worth measuring, or splits attention evenly across every engine, it's describing a market that no longer exists. The shift happened in two quarters. Most tracking setups haven't caught up yet.
What actually changed between January and June
Comscore's panel data, drawn from US desktop, mobile, and home router usage, shows the AI market moving in one direction: ChatGPT down 20 points, Gemini up 13, Claude up 9. Chief analytics and technology officer Smriti Sharma called it essential to any brand's growth strategy in this new era, and the underlying numbers back her up.
Claude's desktop conversations rose from 1.1 million in October 2025 to 22.3 million in June 2026, a twentyfold increase in eight months. Gemini's desktop conversations grew from 8 million to 89 million over the same year. ChatGPT still leads on raw scale, with 168 million desktop conversations in June and more than a billion active users reported in July 2026, but leading and dominating have stopped meaning the same thing.
Similarweb data reported by Search Engine Journal tells a similar story over a longer window. ChatGPT's share of AI-assistant web visits fell from 76.4% to 52.7% year over year, while Gemini rose from 9% to 27.3% and Claude from 1.6% to 8.9%. StatCounter's referral-share numbers move the same way in miniature: Perplexity's worldwide chatbot referral share dropped from 7.91% in June to 4.31% in August, while Gemini's climbed past 10%. Two independent measurement firms, one direction.
Why equal-weighted AI Share of Voice tracking is now misleading
AI Share of Voice only means something if it reflects where your buyers are actually asking questions. A tracker that checks ChatGPT, Gemini, Claude, and Perplexity with equal weight, or checks ChatGPT alone, is answering a question about last year's market. The fix isn't switching to one engine. It's reweighting the ones you already track.
The calculation behind AI Share of Voice hasn't changed: it's still the share of relevant AI answers that mention your brand, measured across a defined set of prompts, as laid out in the original definition and calculation framework. What changed is which engines belong in that prompt set and how heavily each one should count. A measurement methodology built around 2025's usage split is now over-sampling a shrinking engine and under-sampling two growing ones.
Comscore is measuring prompt volume, not the Citation Share that VizibleAI tracks: how often an engine actually cites a brand inside the answers it gives. The two move together most of the time, but not always. A brand can hold steady citation share inside ChatGPT while its overall relevance still shrinks, simply because ChatGPT's slice of total AI queries keeps getting smaller.
A three-tier way to weight your tracking
Search Engine Journal's Greg Jarboe reviewed the same market-share numbers and landed on a practical rule: don't remove Perplexity from your tracker, remove it from equal weighting. That advice generalizes into a three-tier structure that matches tracking effort to actual usage, rather than treating four engines as interchangeable.
Tier 1: ChatGPT and Gemini. Together they carried roughly 80% of US AI prompt volume in June 2026. Track both daily and weight them heaviest.
Tier 2: Claude. An 11% share that grew more than fivefold in six months earns closer attention than its current size suggests. Track weekly, and watch the growth rate rather than just the total.
Tier 3: Perplexity, Grok, DeepSeek, and Copilot. Each holds a low single-digit share of prompt volume, but a brand with a technical or research-heavy audience can still lose real visibility here, the same blind spot this blog has flagged with Mistral and DeepSeek before. Track monthly, or whenever a specific segment skews toward them.
None of this requires adding new engines to your tracker. It means spending the attention you already have where the prompts actually are, and revisiting that split as the market keeps moving.
AI Share of Voice is a moving target twice over. Your brand's mentions can shift week to week, and the engines you're measuring against can shift entirely within two quarters.
What usage share doesn't tell you
Prompt volume measures where the questions are, not where the money is. Anthropic's enterprise base passed 100,000 organizations running Claude on Amazon Bedrock alone in 2026, and a single B2B software buyer researching vendors through Claude can be worth more to a SaaS brand than a hundred casual ChatGPT users asking about dinner recipes.
Two caveats matter before anyone rebuilds a scorecard around this data. Market share doesn't equal pipeline value: AI referral traffic already converts at meaningfully different rates than traditional organic search, so a smaller engine used by high-intent buyers can outperform its market-share ranking on the numbers that end in revenue. And usage share is a national average. A brand selling to developers or researchers should weight Claude and Perplexity higher than that average suggests, not lower.
How often to recheck the weights
Comscore publishes this data quarterly, and that's a reasonable cadence for revisiting engine weights, too: often enough to catch a shift the size of Claude's, rarely enough that normal week-to-week noise doesn't send you chasing a trend that isn't real.
Run the comparison the same way each quarter. Pull your own AI Share of Voice numbers per engine using a repeatable competitive analysis, line them up against the latest usage data, and adjust the weighting rather than rebuilding the whole scorecard. That habit is a much smaller job than starting over, and it's the difference between a benchmark score that describes last year's market and one that describes this one.
Frequently asked questions
What is AI Share of Voice?
AI Share of Voice measures how often your brand is mentioned in AI-generated answers to relevant prompts, compared to competitors, across engines like ChatGPT, Gemini, Claude, and Perplexity. It's the AI-era equivalent of traditional share of voice, tracked inside chatbot answers instead of media impressions.
Why did ChatGPT's AI market share drop in 2026?
Comscore's Q2 2026 AI Intelligence Report attributes the drop to consumers actively diversifying which AI assistant they use as Gemini and Claude improved and expanded distribution, including Gemini's reach across Google's own products and Claude's growth inside enterprise tools built on Amazon Bedrock.
Should I stop tracking Perplexity for AI Share of Voice?
No. Perplexity's overall usage share has fallen to a low single-digit percentage, but dropping it entirely means missing the specific audiences, often researchers and technical buyers, who still use it heavily. Track it less frequently and weight it lower, rather than removing it.
How often should I reweight AI Share of Voice tracking by engine?
Quarterly is a reasonable cadence, matching how often major measurement firms like Comscore publish updated engine usage data. Reweighting more often risks reacting to normal day-to-day measurement noise rather than a real shift in where your buyers are asking questions.
Does a bigger AI market share mean an engine matters more for my brand?
Not necessarily. Market share reflects prompt volume, not buyer intent or conversion value. A smaller engine used heavily by your specific buyers, such as Claude among enterprise software researchers, can matter more to your pipeline than a larger engine used mostly for casual questions.