Google Gemini seems to have lost much of its momentum in recent months — and now even seems to be going back on its initial progress.
Similarweb’s July 2026 ranking of the world’s top-visited websites has Gemini sitting at rank 16, down four places from where it stood in May. ChatGPT, by contrast, held firm at rank 5, unchanged over the same two-month stretch. Claude moved up one spot to 33, DeepSeek dropped 13 places to land at 92, Grok slid 8 places to 150, and Perplexity had the roughest showing of the lot, tumbling 61 places to rank 345. Interestingly, Google had just announced yesterday that Gemini had touched 1 billion monthly users.
Placed side by side, the numbers describe a market where the gap between the leader and the chasing pack has stopped narrowing, and where the site most people expected to keep closing that gap is instead sliding the other way.
Gemini’s Slide Isn’t New, Just More Visible Now #
This ranking drop lines up closely with what Similarweb’s traffic-share data has already been showing for the past couple of months. Gemini had spent over a year on an almost uninterrupted run, climbing from 9.4% of generative AI traffic in July 2025 to over 27% by earlier this year, at one point growing multiple times faster than ChatGPT on a year-over-year basis. That run has now broken. Gemini’s traffic share actually dipped in the most recent monthly reading, the first time it had moved backward since the middle of last year, and the ranking slip to 16th is essentially that same trend showing up in a different dataset.
Much of Gemini’s earlier surge traced back to two specific product moments: the viral launch of the Nano Banana image model, which briefly pushed the Gemini app to the top of the App Store ahead of ChatGPT, and the release of Gemini 3 Pro, which closed a good chunk of the reasoning and coding gap with OpenAI. Nano Banana Pro followed in November, adding 4K image output on top of the same reasoning core, and for a while it looked like Google had built enough of a product lead to keep the traffic gains coming.
What Google hasn’t managed to deliver since is a genuine follow-up. Sundar Pichai told developers at Google I/O in May that Gemini 3.5 Pro was on its way and would ship the following month. It didn’t. Google spent July retraining the model after it reportedly missed internal coding benchmarks, and when that retraining also came up short, the company put out Gemini 3.6 Flash as a stand-in for the flagship release everyone had been told to expect. The reception was lukewarm at best — Gemini 3.6 Flash scored identically to its predecessor on the Artificial Analysis Intelligence Index, and a Semi Analysis report later claimed that Google has quietly shelved 3.5 Pro entirely in favor of pushing resources toward Gemini 4. Rivals noticed the gap too, with executives at OpenAI and Meta publicly mocking Google once Gemini 4’s pretraining run was announced with no ship date attached for 3.5 Pro.
Without a flagship release to lean on, Gemini has spent the past few months running on whatever momentum Nano Banana and Gemini 3 Pro built up last year, and both the traffic-share and website-ranking data are now reflecting that the fumes are running out.
Claude’s Rise Shows Up #
Claude climbing to 33rd, its second consecutive improvement, fits a pattern that has been building across nearly every dataset tracking the AI race this year, not just Similarweb’s. Anthropic’s assistant was sitting at under 2% of generative AI traffic a year ago and has since climbed to nearly 10%, the sharpest sustained gain of any platform tracked. Ramp’s payments data has separately shown Anthropic’s share of combined OpenAI-Anthropic business spend jumping from around 10% at the start of 2025 to over 65% by this February, and Menlo Ventures has reported Anthropic overtaking OpenAI in enterprise LLM API usage.
Pangram, an AI text detection company that tracks which model family produced a given piece of writing, has picked up a similar signal through an entirely different methodology. Its data, drawn from anonymized samples since mid-2024, shows Anthropic’s share of AI-generated text climbing steadily from 4.3% to 14.9% by July 2026, a gradual rise without a single sharp spike that would suggest one viral moment driving it. The same Pangram report found Google’s share falling even more steeply, from around 12% in its early tracking period to just 1.9% by last month, a far lower number than any traffic dataset would suggest, but directionally telling the same story of Gemini losing ground rather than gaining it.
Some of Claude’s recent surge has coincided with unusual circumstances, including the attention around Anthropic’s public standoff with the US government over military use of its models, which briefly sent Claude to the top of the App Store ahead of both ChatGPT and Gemini. But the broader trend predates that episode. Claude’s app DAUs had already grown roughly 5x between October and February, and Anthropic’s revenue run rate has reportedly crossed $44 billion this year.
What Stands Out #
ChatGPT not moving at all between May and July is arguably as notable as any of the individual drops. After more than a year of steadily giving up share to Gemini and Claude, OpenAI’s flagship appears to have found a floor, at least for now, while its two closest challengers are heading in opposite directions — one still climbing, the other sliding back toward where it started.
DeepSeek and Grok’s declines fit a pattern that has held for most of the year, with both platforms drifting without a clear breakout moment since their respective viral launches faded from memory. Perplexity’s 61-place fall is the steepest of the group by a wide margin, and stands out even against a backdrop where most challengers outside the top three are struggling to hold their ground.
Whether Gemini’s slide is a temporary lull while Google waits on 3.5 Pro or Gemini 4, or the start of a more durable shift in the consumer AI market, is a question that will likely take another month or two of data to answer clearly.