ChatGPT market share fell to 50 per cent. Here is where your content should live next.

What Comscore actually measured#

The fox does not judge a hillside by one footprint. On 22 September, Comscore published its Q2 2026 AI Intelligence Report, and Digiday's write-up says ChatGPT's share of prompt volume "fell from 70% to 50% between January and June 2026". Gemini's share "nearly doubled from 17% to 30%", and Claude "climbed from 2% to 11%".

The method matters as much as the headline. Digiday reports that Comscore "used its opt-in panel data to conduct the research, with a respondent base running between 500,000 and 1 million users". That is a large panel, but it is a panel: people who agreed to be measured, counted by prompts sent. It is not a census of every assistant on earth, and it does not measure how many of those prompts were about your category.

So when someone asks whether ChatGPT is losing market share, the accurate answer is that its share of prompts in this panel fell sharply. That is a smaller claim than ChatGPT losing users, and the numbers show why. Advanced Television's summary notes that ChatGPT still reached "168 million desktop conversations in June 2026" and grew its multiplatform visitation "to 99 million from 59 million a year earlier". Share fell because the pie grew and other assistants took bigger slices of it. ChatGPT market share is a slice of a growing pie, so a falling ChatGPT market share is not the same thing as a shrinking ChatGPT.

Usage is expanding on both desktop and mobile. Comscore's data, as Digiday relays it, shows 35% of desktop users visiting an AI assistant in June 2026, up from 25% in January 2025, with mobile reach rising from 12% to 29%. Google desktop searches carrying an AI overview reached 39.4% in June 2026. More people are asking, more of them are asking more than one assistant, and that is the real change for anyone who publishes.

Why ChatGPT market share depends on the ruler#

Here is the part the headline skips. Each ruler measures something different, so each gives a different answer. Comscore counts prompts. Similarweb counts web visits, and says ChatGPT's share of worldwide generative AI web traffic slid "from roughly 76% a year ago to around 53% today". StatCounter counts referral clicks to a million-plus websites, and its April 2026 release, headed by the line "ChatGPT's share has now fallen for three consecutive months to a record low", still put ChatGPT at 76.85% of AI chatbot referrals.

That spread is a gift to a content strategist, because it tells you which ruler to trust for which decision. If you are deciding how much to invest in being cited by a given assistant, prompts show where curiosity is heading. If you are forecasting sessions on your own site, referral share is the ruler that pays the bills, and today it says ChatGPT still sends most of the clicks.

The fourth ruler is reach. The Pew Research Center's 2026 survey of US adults found that "about half of U.S. adults now use AI chatbots". That is ever-used reach, not share of prompts, and it is a reminder that a habit forming across a population is different from a market being divided among vendors.

The direction is consistent even when the levels are not. Each measure shows ChatGPT market share sliding and rivals rising, and each is a slightly different quarry. Only the size of the slide and the size of the starting lead differ. A sensible plan works with the direction and treats every level as provisional.

Gemini market share and Claude market share: who is gaining, and by how much#

Gemini is the clearest winner in the data, which is why ChatGPT vs Gemini is a comparison content teams are increasingly asked about. Comscore's move from 17% to 30% is the fastest gain, and Alphabet's own second-quarter filing says "the Gemini App has 950 million monthly active users" (Alphabet Q2 2026 exhibit). Gemini market share matters doubly for content teams because it sits on the same company as Search, so the choices you make for AI Overviews and for the Gemini app overlap heavily.

Claude market share is smaller but steeper. From 2% to 11% in six months is a fivefold rise on a small base, and Comscore's own figures show volatility: Claude's desktop conversations peaked near 40 million in April before settling at 22.3 million in June, according to Digiday's write-up. A base that small can double or halve on one product launch. Treat Claude's line as a trend to watch, not a number to plan a year around.

That post captures the mood, and it is close to the useful reading, with one refinement. A discovery surface is a place where a stranger meets your answer. Fragmentation means there are more such places, which means more chances to be present and more ways to be absent.

The last piece of context is the oldest. Search engines are not going away. The Reuters Institute's Digital News Report 2026 says weekly use of AI chatbots for news "increased year-on-year from 7% to 10%". Growth is real and still modest. AI Overviews reaching nearly four in ten desktop searches in Comscore's data is the bigger bridge, because it puts an assistant answer in front of people who never chose to open an assistant.

Five content moves when AI search market share is splitting#

None of this argues for a rewrite per engine, or for abandoning an assistant because its ChatGPT market share dipped. Google says there are "no additional requirements to appear in AI Overviews or AI Mode", and that you do not "need to create new machine readable files, AI text files, or markup to appear in these features" (Google Search Central). The same discipline that earns a place in Google's answers, clear, well-sourced pages, is the base for every other assistant too. What changes is where you spend your attention.

The reason to be cautious about single-engine tactics is measurable. An Ahrefs study of 15,000 prompts from 2025 found "only 12% of links cited by ChatGPT, Gemini, and Copilot appear in Google's top 10 results for the same prompt". That study is over a year old and covers long-tail queries, so it is a warning about direction rather than a current constant, but it explains why ranking well in one place does not guarantee a citation in another.

The second move needs a warning. Assistant traffic is under-counted. Seer Interactive's testing found that when ChatGPT links "don't include UTM parameters or referral data", the traffic "will likely appear as" direct in GA4 (Seer Interactive). Treat every AI-referral number you see, including your own, as a floor. The five instruments that measured AI search traffic and none of them agreeing is a useful companion read.

If you are building this for a brand, the practical brief is short. Decide the ten questions your buyers ask before they buy. Answer each in one plain paragraph with a named source. Publish it somewhere you own. Then measure which assistants pick it up. Rank for the question, then earn the citation. That is the work our content marketing team and our search and GEO team do, and it is the same reasoning behind the note on content distribution after the referral collapse.

Pick one page this week and rewrite its opening as a forty-word answer with a named source. Then check, next month, which assistant quoted it. A shrinking ChatGPT market share is a reason to widen the trail, not to leave the thicket. The fox does not need the whole forest mapped to choose the next path, only an honest reading of the one in front of it.

Frequently asked questions#