In the news : On 4 September 2026, Search Engine Land published an analysis of Semrush Traffic Analytics data comparing the audience of the largest US websites in July 2026 against July 2025. Bing is down 50.43% year over year, while ChatGPT is up 48.38% and reaches 1.09 billion monthly visits, ranking 9th among all US websites (Search Engine Land, 4 September 2026).
The number making the rounds since Friday is ChatGPT’s. The one that should concern you is Bing’s.
For three years, the standard agency advice fit in one sentence: “don’t forget Bing, it’s the index behind ChatGPT.” That sentence just lost both of its legs at once. Bing’s audience has halved, and Bing’s index is no longer what decides your presence in ChatGPT.
What the numbers actually say
The data comes from Semrush Traffic Analytics, covering US websites, year over year: July 2026 against July 2025.
- Bing: down 50.43% year over year.
- ChatGPT: up 48.38%, at 1.09 billion monthly visits, 9th biggest US site.
- Google: up 10.72%, still first with 25.3 billion visits.
- DuckDuckGo: down 15.83%.
- YouTube: up 36.6%, at 10.3 billion visits. Reddit stays 3rd with 2.61 billion.
- LinkedIn: up 15.48%, at 473.7 million visits.
- On the losing side: Amazon down 15.31%, Temu down 28.55%, ESPN down 16.0% (366.0 million visits), NBCNews.com down 20.2%.
Two readings emerge. The first is that Google is not collapsing: it is growing. The “Google is dead” narrative does not survive these figures. The second is that the second Western search engine is emptying out. It is not search engines that are losing. It is the second search engine.
Why “do Bing SEO to show up in ChatGPT” no longer holds
That advice rested on a technical reality from 2024: ChatGPT Search then relied heavily on Bing results. Since then, OpenAI has been running its own discovery crawler, OAI-SearchBot, the crawler that indexes the web to feed ChatGPT Search answers, distinct from GPTBot which serves training, and from ChatGPT-User which fetches a page on a user’s request. Allowing GPTBot in your robots.txt is not enough to appear in ChatGPT Search answers: these are three separate mechanisms.
In other words, even if Bing had kept its audience, it would no longer be the gateway into ChatGPT. And since it lost that audience, it is no longer a serious acquisition channel either. Both reasons to invest in it disappeared at the same time.
The nuance most write-ups skip: Bing collapsing as a destination is not Bing disappearing as infrastructure. The Bing index still feeds Copilot on the Microsoft side. These figures measure human visits to bing.com, not third-party product usage of the index. Conflating the two would replace one error with another. The topic connects to the retirement of the legacy Bing APIs at the end of August 2026, which already reshuffled programmatic access to results.
The trap: 1.09 billion visits are not 1.09 billion clicks to you
This is the costliest confusion in the dataset. The figure measures visits to chatgpt.com. It says nothing about outbound clicks to third-party sites. A conversational platform can double its audience while sending proportionally fewer visitors onward, because the answer is delivered inside the interface. That is in fact the expected behaviour.
A measurement problem compounds it. In an analysis published on 5 September 2026, Scalevise notes that referral information does not reliably survive the transition from an AI surface to a website. The result: a share of assistant-driven visits lands in GA4’s “Direct” channel. You underestimate AI at the exact moment it matters most, and you cannot tell which path converts. We already documented this gap between citations and actually attributed traffic.
What to do this week
- Take Bing out of your acquisition targets. Keep Bing Webmaster Tools for indexation checks, a few minutes a month. Remove any traffic goal attached to it: you would be steering a channel that lost half its audience in a year.
- Read your robots.txt line by line. Is OAI-SearchBot allowed? What about PerplexityBot, ClaudeBot, Google-Extended? This is a trade-off, not an obvious call: allowing them buys visibility by giving up content, refusing does the reverse. The debate crystallised around Cloudflare’s default-blocking announcement. What is not defensible is inheriting a default configuration you never chose.
- Segment your Direct traffic. Build a GA4 segment for deep landing pages arriving with no referrer. Nobody spontaneously lands on a three-level URL by typing it. It is not an exact measurement, it is an order of magnitude, and it beats nothing.
- Measure citations, not just clicks. On interfaces that answer without sending the user out, being cited becomes a metric in its own right. The volatility of ChatGPT citations observed on Reddit shows how fast that signal moves: track it regularly, not once a quarter.
The limits of these numbers
This data consists of clickstream panel estimates, not server logs. Semrush does not publish the detailed composition of its panel, and a 50% swing over twelve months can partly reflect a change in sample as much as a change in behaviour. No methodology revision was announced over the period, but nothing rules it out either.
The scope is American. It says nothing about France, where Bing retains different usage shares, particularly in Windows business environments. Finally, one year is a short window on which to settle a structural trend. What these figures establish solidly is an order of magnitude and a direction of travel, not a forecast.
The Cicero take
The real lesson is not “Bing is dead”. It is that search has concentrated on two interfaces, Google and ChatGPT, and that diversification strategies built on the third player no longer have an object. The budget that used to go to Bing should not move to another search engine. It should move to measuring what the assistants say about you.
Method note and disclosure. We confirmed the distinction between OAI-SearchBot, GPTBot and ChatGPT-User in OpenAI’s crawler documentation before publishing. We could not, however, cross-check the Semrush estimates against a second independent measurement source: they are reported as published. Cicero has no commercial relationship with Semrush, Microsoft or OpenAI, and no link in this article is sponsored or affiliated.
Sources
- → Search Engine Land, 4 September 2026: analysis of Semrush Traffic Analytics data, July 2026 against July 2025, US websites. Bing at -50.43%, ChatGPT at +48.38% and 1.09 billion monthly visits (9th), Google at +10.72% and 25.3 billion visits, YouTube at +36.6%, LinkedIn at +15.48%, DuckDuckGo at -15.83%, Amazon at -15.31%.
- → Semrush Traffic Analytics: primary source of the audience estimates quoted, clickstream panel methodology.
- → Scalevise, 5 September 2026: analysis of the measurement gap, loss of referral information between an AI surface and the destination site, consequences for GA4 channels.
- → Search Engine Roundtable: documentation of OpenAI crawlers, distinguishing OAI-SearchBot (discovery for ChatGPT Search), GPTBot (training) and ChatGPT-User (on-demand fetch).
Frequently asked questions
Should you still optimise for Bing in 2026?
Yes, but no longer for the same reasons. The Semrush figures relayed by Search Engine Land on 4 September 2026 show Bing’s US audience down 50.43% year over year: as a direct acquisition channel, Bing no longer deserves a line in your targets. Bing Webmaster Tools remains useful for checking indexation, and the Bing index still feeds Copilot on the Microsoft side. Keeping it costs a few minutes a month. Setting traffic goals against it no longer makes sense.
Is ranking on Bing enough to appear in ChatGPT?
No. OpenAI runs a dedicated discovery crawler for ChatGPT Search, OAI-SearchBot, distinct from GPTBot which is used for training. Allowing GPTBot in your robots.txt does not guarantee you appear in ChatGPT Search answers, and neither does ranking well on Bing. These are three separate mechanisms and they need to be handled separately.
Does 1.09 billion visits to ChatGPT mean more traffic for my site?
Not mechanically. The figure measures visits to chatgpt.com, not outbound clicks to third-party sites. A platform can grow its audience while sending proportionally fewer visitors onward, because the answer is delivered inside the interface. Reading ChatGPT’s growth as a promise of inbound traffic is the most common misinterpretation of this dataset.
Do these numbers apply to the French or European market?
Nothing allows us to claim that. The dataset covers US websites, July 2026 against July 2025. Bing’s market share and AI assistant usage patterns differ noticeably in France. The underlying trend, audience concentrating on conversational interfaces, is plausible here too, but the percentages cannot be transposed as they are.
Growth and SEO & GEO content strategist, I founded Cicéro to help businesses build lasting organic visibility : on Google and in AI-generated answers alike. Every piece of content we produce is designed to convert, not just to exist.
LinkedIn