Discussion
  • AI Search

ChatGPT search partly relies on third-party search providers. How much of ChatGPT visibility work is really Bing SEO in your experience?

Published 1 reply

Discussion starter: an AI search question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.

Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.

1 reply

  • admin
    Admin
    Original poster

    Less of it than the “ChatGPT is just Bing” shorthand suggests, and nobody outside OpenAI can put a reliable percentage on it. Bing work covers one layer: making sure your pages can be found by at least one of the search systems ChatGPT retrieves from. It does not decide whether ChatGPT cites your page once it is retrieved, what the model already believes about your brand, or what reaches ChatGPT through other providers and OpenAI’s own crawler. Treat Bing indexing as a cheap prerequisite to check, not as the main lever.

    OpenAI’s crawler documentation separates GPTBot (training) from OAI-SearchBot (ChatGPT search). Its ChatGPT search help page says ChatGPT search “sometimes partners with other search providers” and points users to Microsoft’s and Shopify’s privacy policies, but it does not say how often each provider is used or for which queries. The public evidence since launch points to several sources rather than one, which makes Bing rankings a risky proxy for ChatGPT visibility.

    What is documented, what is reported, and what is inferred

    Claim Source Evidence level
    ChatGPT search sometimes uses partner search providers and typically rewrites a prompt into targeted queries for them; OAI-SearchBot crawls for ChatGPT search and GPTBot for training OpenAI ChatGPT search help page and crawler documentation Platform documentation
    OpenAI reportedly used, and may still use, a Google-scraping service (SerpApi) for answers on real-time topics such as news, sports and finance Report by The Information, summarised by Search Engine Land in August 2025; the article does not report OpenAI confirming or denying it Press report
    A ChatGPT executive indicated during Google’s antitrust trial that Bing had significant quality issues and that Google’s index would accelerate OpenAI’s search ambitions Same Search Engine Land article Reported testimony
    ChatGPT’s streamed response data from 21 May to 21 July 2026 carried a result_source field with four values (Labrador, Bright, Oxylabs, SERP). The author reads Labrador as OpenAI’s own index with vertical sub-indexes such as news, local and shopping, and the other three as external scraping providers. He concludes that ChatGPT still uses both Microsoft and Google, that Bing shows up mainly in Deep Research, and that the overlap between Bing visibility and ChatGPT visibility is lower than most people assume Peec AI analysis, September 2026 Third-party analysis by an AI visibility vendor; not confirmed by OpenAI

    The Search Engine Land report and the Peec AI analysis support a limited conclusion: ChatGPT retrieval draws on more than one source, and Bing is one input rather than the input. They do not tell you the mix for your queries, and the mix may differ by topic. The SerpApi report concerned real-time topics, and the Peec analysis describes separate vertical indexes. Field names in streamed data can also change without notice, so treat the specific labels as a snapshot, not a stable interface.

    One detail in the Peec analysis cuts the other way. It suggests that results which look like Bing may be routed through Web IQ, Microsoft’s grounding service for AI agents, which is built on the Bing index. If that is right, being indexed in Bing still matters for the Microsoft route even where Bing’s rankings do not.

    Four layers of ChatGPT visibility, and how much of each is Bing

    It helps to split “ChatGPT visibility” into the stages an answer passes through. OpenAI’s help page says that when ChatGPT search uses a partner provider, it “typically rewrites your query into one or more targeted queries” before sending them. That makes rankings for those sub-queries likely to matter more than rankings for the prompt as typed.

    Layer What decides it How much is Bing work What to do
    Access OAI-SearchBot and the providers’ crawlers can fetch your pages; nothing blocks them in robots.txt, the CDN or the WAF Partly: Bingbot access and Bing indexing are one piece Allow OAI-SearchBot and Bingbot, and confirm both in server logs
    Retrieval Your pages turn up for the searches ChatGPT runs, across providers and OpenAI’s own index Partly, and only where Bing is the source used Rank for the sub-questions buyers ask, in Bing and Google; keep key pages indexed in both
    Citation Which retrieved page ChatGPT cites for a given statement Not directly Pages that answer the sub-question plainly, with current facts; independent coverage of you
    Memory What the model already knows about your brand from training No Consistent descriptions of what you are, and for whom, across your site and major third-party sources

    There is also a route that is not web search at all. The Peec analysis points to verticals such as local and shopping, and recommends that local businesses keep profiles such as Yelp and TripAdvisor accurate. None of that is Bing SEO, and for a restaurant or a retailer it may matter more than any web ranking.

    Important distinction:

    Bing work can make your page retrievable. It cannot make ChatGPT choose your page over a competitor’s comparison, a review site or a forum thread once all of them have been retrieved. Much of the work that separates you from competitors happens after retrieval.

    The Bing work worth doing anyway

    Most of this is configuration and auditing rather than a content programme, and it also protects your visibility in Bing and Copilot, whatever the ChatGPT share turns out to be.

    • Verify the site in Bing Webmaster Tools, submit sitemaps, and use URL Inspection to confirm your money pages are indexed in Bing, not only in Google.
    • Compare Bing and Google index coverage for your top pages. Pages indexed in Google but missing from Bing are the gap to fix first.
    • Use IndexNow to tell Bing about new and changed URLs, so fresh content is not waiting on a recrawl.
    • Check CDN and bot-management rules. Cloudflare added a “Disallow AI Training” option for mixed-use crawlers in September 2026, and its full “Block” option can remove mixed-use crawlers from search as well. Confirm that Bingbot and OAI-SearchBot receive normal 200 responses.
    • Check server logs for OAI-SearchBot, ChatGPT-User and Bingbot requests to your key pages. If none of them ever fetch a key page, find out why before assuming ChatGPT can see it.
    • Do not read Bing’s AI Performance report as ChatGPT data. Microsoft launched it as a public preview in February 2026, covering citations across Copilot, Bing’s AI summaries and some partner integrations, with no click or traffic data.

    How to estimate your own Bing dependency

    The only honest answer to “how much of it is Bing” is a site-specific one. You can approximate it with a prompt panel.

    1. Build a panel of 30 to 50 prompts that real buyers use, and run each several times in temporary chats, because answers and citations vary between runs. Record every cited URL and the statement it supports.
    2. Write down the likely sub-queries behind each prompt. Some practitioners extract the searches from the data ChatGPT sends to the browser; that is undocumented and can change, so manual rewrites are the fallback.
    3. For each sub-query, record the top 10 results in Bing and in Google.
    4. Classify each citation of your domain, and of the two or three competitors you care about: ranks in Bing only, Google only, both, or neither.
    5. Read the split. Mostly “Bing only” points to Bing-specific work having direct value for you. Mostly “both” suggests Bing and Google SEO are the same work on these topics, and the Bing-specific part shrinks to indexing and access. A large “neither” share suggests retrieval is coming from somewhere else, such as OpenAI’s own index, other providers or vertical sources, and Bing work is unlikely to move it.

    Rank overlap is correlation, not proof of the source. A page that ranks in Bing may still have been retrieved through another route. The split is useful for direction, and it is far better than assuming.

    Hypothetical example:

    A B2B software company runs a 40-prompt panel five times and gets 120 citations of its own domain. Mapping the sub-queries shows that most cited pages rank in both engines, a small group ranks only in Google, very few rank only in Bing, and a sizeable group ranks in neither. For this company, a Bing-specific programme would chase the smallest bucket. The better investment is the “neither” group: finding out which sources ChatGPT is using for those prompts, and whether the company appears in them.

    A fast test: does fixing Bing coverage change citations?

    • Hypothesis: pages indexed in Google but missing from Bing are cited less in ChatGPT, and fixing their Bing indexing raises their citation rate.
    • Test group: 20 to 40 pages missing from Bing’s index, each mapped to prompts in your panel. Control: comparable pages already indexed in both engines, on similar topics.
    • Change: fix access and indexing, then submit through IndexNow and Bing Webmaster Tools.
    • Primary metric: citation rate, meaning the share of panel runs that cite each page. Secondary: OAI-SearchBot fetches of the test pages, and ChatGPT sessions in GA4’s AI Assistant channel group, remembering that visits without a referrer still land in Direct.
    • Window: start the clock when Bing Webmaster Tools confirms indexing, then run the panel weekly for four to six weeks. The change is binary and quick to verify, so a long window adds little; what you need is enough weekly runs to see past the run-to-run variation.
    • Decision: if the test pages’ citation rate rises towards the control group’s, Bing coverage was a real bottleneck, so extend Bing hygiene across the site. If it does not move, the constraint is in citation or in other sources, so shift effort to content that answers the sub-questions and to the third-party pages that get cited in your category.

    Keep the size of the prize in view

    BrightEdge data released in September 2026 put ChatGPT at about 95.1% of tracked AI referral traffic, while AI referrals remained a small share of total referrals compared with search; the methodology is not detailed. That makes ChatGPT the largest AI referrer in that dataset. It does not make a Bing-specific content programme worth funding before your own panel shows that Bing rankings explain your citations.

    My practical position

    Bing SEO is the floor of ChatGPT visibility, not the lever. Make sure Bing and OpenAI’s crawler can reach and index everything that matters, because it is cheap and a gap there removes the page from at least one of ChatGPT’s likely sources. Then measure your own split before moving budget. If your citations track Bing rankings that Google does not share, invest in Bing. If they do not, the work that separates you from competitors is in the citation layer and in the third-party sources ChatGPT already trusts for your category.

    Need help with this on your own site?

    Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.

    See SEO services

Add a reply

Replies are open to approved members. Log in or apply to join.