IndexNow reaches Bing, not every AI engine
Crawlmind Engineering··5 min read
IndexNow is a push protocol that tells participating search engines a URL was added, changed or deleted, and in AI search it reliably reaches only the surfaces built on those engines, which in practice means Microsoft's.
That distinction matters because IndexNow is often described as an "AI freshness" switch, and Microsoft's own messaging links it to generative answers. The protocol itself makes a narrower promise. Before you report a ping as an AI visibility action, it is worth checking who actually receives it.
#Who is on the receiving end
The protocol keeps a public registry of participants. As of this writing, indexnow.org's searchengines.json lists seven: Bing, Yandex, Seznam, Naver, Yep, the Internet Archive and Amazon Bot. The IndexNow specification for search engines requires each participant to share verified URLs with the others, and to do so within 10 seconds of verification. One submission to any endpoint therefore fans out across that whole list.
Three names are missing, and they are the ones most AI visibility programs care about:
- Google. Google said in November 2021 that it would test IndexNow for sustainability reasons. It does not appear in the participant registry today.
- OpenAI. The OpenAI crawler documentation describes OAI-SearchBot as the agent used to surface sites in ChatGPT's search features. It covers robots.txt control and nothing about IndexNow or push submission.
- Perplexity. The Perplexity crawler guide describes PerplexityBot and how to allow it through robots.txt, IP allowlists and WAF rules. It also has no URL submission mechanism.
Anthropic's Claude and the other assistants that run their own retrieval are likewise absent from the registry. A ping does not reach any of them directly.
#Where IndexNow does touch AI answers
Microsoft is explicit about the link on its own surfaces. Its AI Performance announcement for Bing Webmaster Tools describes a report covering citations in Microsoft Copilot, AI-generated summaries in Bing and select partner integrations, and presents IndexNow as the way to keep those answers pointing at the current version of a page. Bing's sitemap guidance for AI-powered search makes the same case, stating that freshness signals affect how quickly updates reach search results and AI-generated answers.
So for Copilot and Bing's own summaries, IndexNow is a first-party, documented input. That is a real and useful claim.
ChatGPT is the contested case. OpenAI's help center says ChatGPT search for Enterprise and Edu may share disassociated queries with Bing to return web results, and its general ChatGPT search article describes search as drawing on third-party search providers plus content from partners. A fresher Bing index can plausibly help there. OpenAI also runs its own crawler, and it has not documented how Bing results and OAI-SearchBot fetches combine, or whether the mix is the same across plans. "IndexNow updates ChatGPT" is an inference, not a documented pipeline. If you write it in a client report, write it as one.
#Google has a push API, but not for you
Google does offer push notification, with tight limits. The Indexing API only accepts pages carrying JobPosting markup or a BroadcastEvent embedded in a VideoObject. Its default quota is 200 publish requests per day, and going beyond that or using it at all for those page types requires an approval form.
Using it for ordinary articles or product pages is outside its documented scope. For AI Overviews and AI Mode, a changed page still reaches Google through normal crawling, internal links and sitemaps. There is no ping to send.
#A 200 means received, not read
Even on participating engines, the protocol's promise is modest. The IndexNow documentation states that the response code only indicates the search engine received the URL. It says nothing about crawling or indexing. The codes worth knowing:
- 200: submitted successfully.
- 202: received, key validation still pending.
- 403: the key was not found or did not match the key file.
- 422: the URL does not belong to the host, or the key does not match the schema.
- 429: too many requests, treated as potential spam.
Two setup errors cause most silent failures. The first is key file scope: a key file placed under a subdirectory only authorizes URLs under that same path, so a key in /catalog/ cannot vouch for /help/. The second is treating a 202 as done. If the key never validates, the batch is dropped without anyone noticing.
Batches can carry up to 10,000 URLs per POST. That limit is generous enough that the temptation is to resubmit the whole site on every deploy. Resist it. Bing's sitemap guidance warns against setting lastmod to the sitemap generation time unless the content actually changed, and the same logic applies to pings. A signal that fires on every build stops meaning anything.
#The Amazon line item
Amazon Bot's presence in the registry is the most interesting entry for AI visibility, because Amazon's crawlers feed Alexa. The Amazonbot documentation lists three agents (Amazonbot, Amzn-SearchBot and Amzn-User) and ties Amzn-SearchBot to search experiences such as Alexa. The page does not mention IndexNow, and Amazon has not described what it does with the URLs it receives through the protocol. Participation is confirmed. Effect on answers is not.
#What to do with this
IndexNow is still worth running. It is cheap, it is documented to help Copilot and Bing, and the fan-out to the other participants costs nothing extra. The mistakes are in how it gets scoped and reported.
- Keep the claim narrow. In reporting, describe IndexNow as a Bing and Copilot freshness input. Do not list it as covering ChatGPT, Perplexity, Claude or Google AI surfaces.
- Ping on real changes only. Trigger submissions from content edits, price changes and deletions, not from deploys. Keep sitemap lastmod honest for the same reason; Bing says it revisits sitemaps at least once per day.
- Log the response codes. Alert on 403, 422 and 429, and on any 202 that never turns into a successful submission.
- Verify the fetch, not the ping. After a submission, look for a verified Bingbot request to that URL in your server logs. For engines outside the registry, the only freshness levers are crawl access and internal linking, so check that OAI-SearchBot and PerplexityBot reach the page with a 200. OpenAI notes that robots.txt changes can take about 24 hours to take effect in its systems, which is a useful floor for how quickly anything changes on that side.
- Use Bing's AI Performance report as the check. It is the one place where a freshness action and AI citations appear in the same first-party dataset. If cited URLs in Copilot still show an old page weeks after an update, the ping was not the bottleneck.
The practical rule: IndexNow shortens the gap between an edit and one family of AI answers. For every other engine, the page still has to be found the slow way.
Related field notes
September 24, 2026 · 6 min
Google pays for grounding, not for links
Google's AI contribution pilot pays when a page shapes an answer, not when it is linked afterward. That rule says a citation count measures the wrong thing.
September 23, 2026 · 5 min
A browser agent is not a crawler
Agentic browsing runs inside the user's own session, so robots.txt, bot allowlists and crawler analytics all miss it entirely.
September 22, 2026 · 4 min
An MCP endpoint is not a discovery channel
NLWeb and MCP make your site answerable by agents that already found you. Nothing on the open web is hunting for a /mcp route yet.
Share or discuss
New posts, no spam. Roughly monthly. Unsubscribe with one click.