Bing, Copilot, and the AI overlap you can actually work.

Bing is the largest measurable overlap in AI search that you can act on directly. Microsoft Copilot builds its web answers on Bing — that one is documented, not inferred. ChatGPT is a different kind of claim: roughly 87% of its live-answer citations match Bing's top results for the same query (Seer Interactive, 2026), a correlation independent testers measured from the outside rather than a mechanism OpenAI publishes. Bing is also the one AI-relevant index that accepts submissions: verify your domain, host an IndexNow key file, and every page you publish can reach it within minutes. This page separates the three things most guides blend together — what is documented, what is measured, and what is worth your Tuesday.

What OpenAI actually documents

Start with the primary source, because almost no guide does. OpenAI publishes documentation for its crawlers, and it is specific: OAI-SearchBot is the crawler behind ChatGPT Search — the surface that produces live answers with links. GPTBot is a separate crawler, primarily collecting training data. Blocking GPTBot is a legitimate choice about your content being trained on, and it does not remove you from ChatGPT Search. Blocking OAI-SearchBot does.

What that documentation does notsay is which index or indexes sit behind ChatGPT Search. OpenAI does not name Bing as the index, much less the only one, and ChatGPT Search can work with more than one search provider. So the familiar claim that ChatGPT simply runs on Microsoft's search results is an inference dressed as plumbing. An earlier version of this page made the same mistake. We took it out, and the page got better, because what replaced it is checkable.

What independent testing measures

With the documentation exhausted in one paragraph, everything else about ChatGPT and Bing comes from people testing the surface from outside and counting what they see. The most quoted count is also the most useful:

Two readings fit that number, and honesty means holding both. Either ChatGPT's retrieval leans on an index that closely resembles Bing's, or both systems reward the same pages for the same reasons — crawlable, current, direct, well-linked — and the overlap is a shared taste rather than a shared pipe. Nobody outside OpenAI can settle it from published material. What either reading supports is the same practical conclusion: a page that Bing has never indexed is a page that sits outside the pattern ChatGPT's citations follow.

What the number does not support: treating Bing as a requirement, a switch, or a submission route into ChatGPT. There is no ChatGPT submission portal at all — our guide to getting mentioned by AI spends a section dismantling that myth. If a vendor tells you their Bing work gets you into ChatGPT, they are selling a pipeline nobody has published.

Why the overlap is still worth your Tuesday

Here is the argument for Bing that survives the caveats: it is the only part of this problem you can act on directly. The alternative to working a measured correlation is not working a documented mechanism — no such mechanism is published — it is guessing. Bing gives you a free console, an open submission protocol, and a rank you can read this week with your own eyes. Cheap, fast, checkable.

And nearly unattended. SEO effort is overwhelmingly Google-only, and practitioner write-ups of Bing-side AI visibility (Conbersa and Euskal Conseil among them) keep finding the same pattern: sites tuned exclusively for Google carrying Bing-side gaps nobody is watching — unverified domains, unsubmitted sitemaps, crawl errors. Those gaps hide pages from Copilot outright and keep them out of the pattern ChatGPT's citations follow. Low competition and a large measured overlap is an unusual combination; it is the reason this work is underpriced rather than obvious.

It is also the reason we automate this specific link rather than talking about it — our AEO service submits every approved page to Bing within minutes of publishing and then checks the standing, which is the whole of what anyone can honestly do on this side of the engines.

Copilot: the straightforward half

Everything above is about inference. Copilot needs none. Microsoft built an assistant on Microsoft's search stack, and its web answers draw on Bing's index and answer layer. Rank in Bing and Copilot can use you; stay out of Bing and it cannot. That is the whole model.

Copilot deserves more attention than it gets on its own merits. It ships inside Windows, Edge, and Microsoft 365, so the person asking it for a vendor, a service, or a comparison is often mid-workday — a business buyer, not a browser. ChatGPT is the volume story; Copilot is the context story. And for Copilot, Bing work is not a proxy for anything: it is the work. (For how ChatGPT chooses between its training memory and a live search in the first place, see how ChatGPT recommends businesses.)

The one AI-relevant index that accepts submissions

You cannot submit anything to ChatGPT: no console, no form, no ping. Copilot has no direct portal either. But the index Copilot answers from takes submissions all day, through two channels that have existed for years:

• Bing Webmaster Tools — the one real webmaster console anywhere near the AI answer stack, as Bing-focused practitioners like Stackmatix have been pointing out. Verify your domain, submit your sitemap, and use index inspection to see which pages Bing kept and which it skipped. If you have only ever looked at Google Search Console, this is a half of your visibility you have never seen reported.
• IndexNow — an open submission protocol Bing accepts. Instead of waiting for a crawler to wander back to your site, you ping an endpoint the moment a page is published or changed, and the notification reaches Bing within minutes.

The key file: ten minutes most sites never spend

IndexNow needs proof that whoever is pinging for your domain actually controls it. The proof is a text file, and setting it up is the entire barrier to entry:

1. Generate a key.An IndexNow key is a long alphanumeric string — your tooling can generate one, and Bing's webmaster documentation walks through it.
2. Host it at your domain root. The key lives in a plain-text file named after itself — yourdomain.com/<key>.txt — containing the key as its only content.
3. Ping on every publish. Each submission carries the URL plus the key; Bing checks the key file on your domain to confirm the submitter and the site match.
4. Leave it in place. The file is the standing proof. Take it down and submissions carrying your key stop being verifiable — which means they stop counting.

That is the whole setup: one file, hosted once. What the submissions carry is a separate question — pages still have to be worth indexing and worth quoting, which is the territory of the ChatGPT SEO playbook (answer-first page shapes, open crawler access) and the full AEO playbook (structured data, mentions, measurement). The key file is the distribution step those pages assume already exists.

Domain standing: the needle that moves in days

The outcome you actually want — your name inside AI answers — arrives on the engines' schedule, not yours. Watch citations alone and you fly blind for weeks, unable to tell “working, not landed yet” from “broken at step one.”

The earlier signal is the sequence itself. Each step below is independently checkable, days before any citation could land:

1. Submission out — minutes after publishing, and loggable, so you know the ping happened.
2. Page in the index — days, checked via index inspection or a site:yourdomain.com search on Bing itself.
3. Domain standing on your buyer queries— the leading indicator. Search your target questions on Bing and note where your domain lands. Movement here means Copilot can now find and rank you, and that you have entered the population of pages ChatGPT's citations are drawn from; stalling here tells you the problem is relevance or competition, not plumbing.
4. Copilot answers— ask Copilot your buyer questions and watch whether your pages start informing them. This is the first real answer-layer confirmation, and it comes well before ChatGPT's.
5. ChatGPT citations — the lagging, least controllable confirmation, weeks out.

Reading a stall is as valuable as reading progress. Stuck at step two — submitted but never indexed — points at crawlability or page quality; the six-reason diagnostic covers the usual suspects. Stuck at step three points at the page answering the wrong question, or answering it worse than incumbents. Moving at three but silent at four and five means wait — the churn in AI-cited sources cuts both ways, and the window reopens continuously.

Where the tool market stands on Copilot

For an engine wired this directly into a major index, Copilot gets remarkably thin coverage from AI-visibility tools — worth knowing before you pick one, whatever you pick:

• Trakkr tracks eight models on every paid plan (from $100/month) — and its published engine list includes no Microsoft Copilot in prompt tracking.
• Profound covers ChatGPT only at its $99 Starter tier and three engines at its $399 Growth tier; per its published coverage, Copilot sits among the engines requiring a custom enterprise contract (the full breakdown is in our side-by-side with Profound).
• Otterly.AI includes Copilot among its four base engines on every tier, starting with Lite at $29/month.

Published list prices, verified August 2026. The full vendor-by-vendor table lives in AI visibility tool pricing.

An honest word on when a tracker is the right pick: if your team already handles content, publishing, and indexing, and what you need is Copilot answers watched on a budget, Otterly.AI is a reasonable choice precisely because Copilot ships in its base engine set at the lowest published entry price in the category. The limits of tracker-only tools show up after the number arrives — something still has to build the page, submit it to Bing, and check back. If that something is you, a tracker is enough. If it is not, the number is where your work starts, not ends.

Who publishes this page — and what we sell

AEO Action wrote this guide, and Bing is the link of the loop our product automates most directly. Its Bing Index Engine submits every landing page you approve via IndexNow within minutes of publishing — with the boundary printed above intact: submission is ours, the indexing decision is Bing's, and we show you both. It checks that your IndexNow key file is live on the domain before submitting, and skips rather than fakes the ping when it is not. Domain-standing checks on Bing run on demand at the domain level, so the leading indicator above is a button instead of a manual search ritual. Copilot itself is scanned by reading Bing's answer layer — an engine pick on every paid plan, with Solo picking 3 of 5 from a pool that includes Copilot and Agency running all six — and every approved landing page is re-checked against its target question on following scans for up to 120 days, so a first citation comes with a dated receipt: shipped date A, first cited date B. We sell the submission and the measurement. Nobody can sell you the engines' decision, and we would rather say so on the page than in the fine print.