Google and Bing are the only two Western search engines with a web index of their own: they send out their own crawlers, build their own archive of pages and decide for themselves how to rank them. Everything else, from DuckDuckGo to Ecosia, buys results from one or the other.
The comparison between the two is almost always made on national market share, and it stops there: Google 91.75%, Bing 5.67% in the UK in August 2026 according to StatCounter. It is the least useful number in the discussion. In the same month, in the same market, Bing accounts for 10.38% of desktop searches and 0.48% of mobile ones. More than twenty times the difference, depending on which device the searcher is sitting in front of.
How much Bing matters in the UK, split by device
This table is the only thing you need to know before deciding whether Bing deserves any work. The data is from StatCounter Global Stats, United Kingdom, August 2026, based on a stated sample of more than 3 billion page views a month.
| Search engine | All platforms | Desktop | Mobile |
|---|---|---|---|
| 91.75% | 85.90% | 97.60% | |
| Bing | 5.67% | 10.38% | 0.48% |
| Yahoo! | 1.25% | 2.20% | 0.68% |
| DuckDuckGo | 0.71% | 0.85% | 0.60% |
| Ecosia | 0.30% | 0.36% | 0.30% |
| Yandex | 0.28% | 0.27% | 0.30% |
On mobile Google takes practically everything and the discussion is over. On desktop, about one UK search in seven goes through something that is not Google, and the biggest slice of that share belongs to Microsoft. If your audience searches from their phone, Bing is statistical noise. If they search from a work desk, it is a tenth of the market.
The number that explains the rest is in the browsers. Again on desktop in the UK in August 2026, StatCounter puts Chrome at 56.02%, Edge at 22.18%, Safari at 12.44% and Firefox at 3.98%. Edge ships with Bing as its default search engine, and Bing's desktop share is a little under half of Edge's: a large part of Bing on desktop is the Edge default that nobody changed, on company computers that nobody configures.
Desktop changes the question, not the answer
For a clothing e-commerce site, where mobile traffic regularly goes above 70%, Bing is a box to tick in half a day and then forget. For an industrial manufacturer selling to a purchasing department, the picture is different: whoever searches for "industrial gasket supplier" does it during working hours, from a Windows PC, often inside Edge, and that traffic lands on a tenth of the market that agencies tend to ignore.
The UK context makes this less marginal than it looks. According to the ONS article Artificial intelligence in UK businesses: 2023 to 2026, based on the Business Insights and Conditions Survey, AI use among UK businesses with 10 or more employees rose from around 12% at the end of 2023 to around 35%, with larger firms ahead. Most of that use happens on work computers, not phones, and that is where Bing's remaining share lives.
The working rule is simple and almost nobody applies it before deciding: open the device report for organic sessions over the last twelve months. Above 50% desktop, Bing is worth an hour's work a week. Below 30%, leave it and spend that time on the site's technical SEO.
Bing's index matters more than Bing's SERP
This is where the useful comparison parts ways with the useless one. Bing.com's share of searches says little, because Bing's index is resold and reused in places where nobody sees the Bing logo.
Microsoft retired the old Bing Search APIs on 11 August 2025, moving all customers to Grounding with Bing Search inside Azure AI Agents. It was not a shutdown, it was a change of use: the index is no longer sold to people building search engines, it is sold to people building agents that need to read the web.
Microsoft's documentation says so explicitly. The page on web search privacy and security in Copilot Studio states that, depending on the configuration, "all or a subset of public websites indexed by Bing are searched" through the Grounding with Bing Search APIs. Same story in Dynamics 365 Business Central, where Copilot's web search is described as "powered by Microsoft Bing". And DuckDuckGo, in its help pages, writes that traditional links and images come "largely from Bing", alongside its own crawler, DuckDuckBot.
In plain terms: every time an employee opens Copilot inside a Microsoft environment and asks a question that needs the web, the answer comes from a subset of pages indexed by Bing. If your site is not in that index, you do not exist in that conversation, and you will not notice because it leaves no trace in Search Console.
"ChatGPT runs on Bing" is no longer a valid answer
It is the most repeated line in British and American guides on the subject, usually brought out as the final argument to convince someone to do SEO for Bing. It was true in 2023 and it is no longer true in the same way.
On 22 April 2025, testifying in the antitrust trial against Google, Nick Turley, head of ChatGPT at OpenAI, confirmed that ChatGPT drew on Bing's search results and spoke of "significant quality issues" with what he referred to only as "Provider No. 1". In the same hearing he said OpenAI was building its own index and hoped to serve 80% of queries from it by the end of 2025, while admitting that the goal had slipped by years. An account of the hearing is on Courthouse News Service.
That the switch is under way was documented by Peec AI, which observed ChatGPT's server-side events between May and September 2026: a result_source field returning four values, "Labrador" (OpenAI's own index), "Bright", "Oxylabs" and "SERP", and an A/B experiment called prefer-index-over-serp-v3 running on about 8% of traffic in mid-August 2026. It is a third-party observation, not an official OpenAI statement, and it does not quantify how much of today's answers come from each source. But the direction is clear, and it matches what Turley said under oath a year earlier.
The practical consequence: optimising for Bing in the hope of showing up inside ChatGPT is a bet on infrastructure that is shifting under your feet. If the goal is visibility inside assistants, the route is different and runs through how models cite their sources, a subject we cover separately in the guide to generative engine optimisation and in the one on how to appear in ChatGPT answers. On Copilot, however, the "it goes through Bing's index" reasoning still holds, because there Microsoft's documentation puts it in writing.
Where the two engines really diverge
Bing has done something Google never has: it has published a list of its ranking criteria. The Webmaster Guidelines, in the version updated in June 2020 and covered at the time by Search Engine Land, name six: relevance to the intent of the query, quality and credibility of the page and of its author, user engagement (clicks, dwell time, query reformulation), content freshness, location of the user and the server, and page load time.
Google publishes nothing similar and sticks to qualitative guidelines. This asymmetry is the real reason it is worth knowing Bing even if you do not spend a penny on it: its stated criteria show which signals an engine with less behavioural data than Google is forced to use more directly. Two of the six concern user behaviour and freshness, and they are exactly the ranking factors the industry tends to treat as folklore.
The visible effect on a real site: pages with vague titles and badly built URLs drop on Bing before they drop on Google, because the match between the query and what the page says it is carries weight more literally. If you already take care of search intent and title tags for Google, the Bing part is already done.
What changes in practice on your site
Almost nothing on the content, something on the tools. There is no "Bing version" of the text: writing the same page twice is the fastest way to create duplicate content and fix nothing on either engine.
| Task | Bing | |
|---|---|---|
| Site verification | Search Console | Bing Webmaster Tools, with direct import of properties from Search Console |
| Submitting new or updated URLs | XML sitemap and crawling | XML sitemap plus IndexNow: instant notification, up to 10,000 URLs per single POST request |
| Who receives the IndexNow signal | Nobody: Google does not take part in the protocol | Bing, Yandex, Naver, Seznam.cz, Yep and Amazon |
| Structured data | schema.org | schema.org, same markup |
| Published ranking criteria | No | Yes, six factors stated in the Webmaster Guidelines |
| Local listing | Google Business Profile | Bing Places for Business |
On IndexNow it is worth being precise, because it is the most concrete technical difference between the two worlds and it is often described badly. You generate a key of 8 to 128 characters, upload a text file with that name to the root of the domain and send the changed URLs through the API. The official documentation is also clear about the limits: no guarantee of indexing, every URL submitted uses up the site's crawl quota, and you should wait at least five minutes before resubmitting the same address. It is there to say "this page has changed now", not to get you ranked.
For a publishing site or an e-commerce store with prices that move every day, it is half an hour of setup that shortens update times on six engines. For a brochure site updated three times a year, it changes nothing.
The paid side: Google Ads and Microsoft Advertising
On campaigns the comparison flips compared with organic, because costs come into play. The claim that circulates in every article on the subject is that cost per click on Microsoft Advertising is noticeably lower: you find it repeated with different percentages (up to 30%, up to 50%) and almost never with a primary source behind it. We do not report it as data because it cannot be checked against an official document, and average CPCs only make sense within a specific sector and market.
What can be said without inventing numbers is structural. The pool is smaller, in proportion to Bing's desktop share, so the volume of conversions you can get is capped whatever the budget; auction competition on many B2B queries is lower simply because half the agencies never open the account; and company targeting based on LinkedIn's professional data exists only inside Microsoft Advertising. The right way to test it is to duplicate a campaign that is already mature on Google Ads, let it run for a month on a minimal budget and compare cost per acquisition, not cost per click. If you need a hand setting up the test, it is one of the things we do in our Google Ads services; on the organic side, the reasoning extends to assistants in our AI SEO services.
We have collected the numbers on Google Ads statistics and on the benefits of Google Ads separately, so as not to mix them with the organic figures on this page.
First-hand data on the two ecosystems
We run continuous AI visibility monitoring for an Italian LED wall manufacturer that sells across Italy. In the four weeks to 13 September 2026 the brand was searched with the same prompts on two generative engines, four runs a week on each. The detection rate on ChatGPT went from 75% to 50%, the one on Gemini from 75% to 87.5%, over the same period and with the same website.
The sample is small and does not prove a market trend: four weekly runs per engine are a thermometer, not a study. It does, however, say something that bears directly on the Google versus Bing comparison: generative engines diverge from each other more than the two classic SERPs do, and they diverge within weeks. If you measure your presence on a single assistant, you are looking at one slice of something that moves at different speeds depending on the index underneath. We have written about how we measure these numbers on an industrial case in the piece on SEO for manufacturing.
When working on Bing pays off and when it does not
It pays off when more than half of your organic traffic is desktop, when you sell to businesses that work inside the Microsoft ecosystem, when the site publishes or updates content every day and IndexNow has something to notify, and when you are in a sector where competitors have not even verified their domain in Bing Webmaster Tools.
It does not pay off when traffic is mostly mobile, when the site still has open crawling or page speed issues on Google, when the keyword research has not been done or when the main pages do not yet have a decent structure. In all these cases, working on Bing is an elegant way to put off what actually needs doing.
There is a third category, the one most businesses fall into: the site is in order, traffic is mixed, and the question "Google or Bing" is simply the wrong one. They are not two alternative strategies, they are two indexes reading the same site. There is more on this in the comparison of alternative search engines to Google and in the piece on how search engines work.
One thing to do this week, which needs no budget and does not appear in any guide comparing the two engines: open your server logs and count how many times Googlebot, bingbot and OAI-SearchBot have visited over the last thirty days. The ratio between those three numbers tells you how much your site really exists for each index, and it almost never matches the market share you find in the tables. It is the only Google versus Bing comparison that talks about your site rather than the market in general.
Frequently asked questions
