This is a write-up of something that happened to us, on our own site. We're publishing the full trail — including the part where we admit we'd never checked — because the blind spot we found is almost certainly not unique to us.
The setup
We were testing AnythingLLM — a local LLM stack with chat, RAG, and web search — as part of our ongoing work on how machines read the web. Local setups like this are worth watching precisely because they reveal the defaults: what a tool does when nobody configures anything.
One of those defaults turned out to matter more than we expected. AnythingLLM offers several web search providers. Nearly all of them — Perplexity's API among others — require an API key and a paid plan. One doesn't: DuckDuckGo. So DuckDuckGo is what you use, and DuckDuckGo is what we suspect a large share of local and budget AI deployments quietly use, for exactly the same reason we did. Zero friction wins.
Then we asked the assistant to analyze one of our own sites.
It replied that the site didn't exist in search results.
The investigation
First hypothesis: tool error. We re-ran the query. Same answer. So we cut the tool out of the loop and searched DuckDuckGo directly, by hand.
Nothing. Not ranked low — absent. A site that has ranked on the first page of Google for its core query, for years, returned zero results on DuckDuckGo.
Second step: where does DuckDuckGo get its index? It doesn't maintain a full web index of its own. Its organic results draw primarily on Microsoft's Bing index, supplemented by its own crawler and other sources. The same is true, in various configurations, for Yahoo and Ecosia. Bing's index is, in practice, the second backbone of the searchable web — and the default backbone for a lot of tooling that can't or won't pay for anything better.
Third step: check the source. We opened Bing Webmaster Tools and looked for the site in the index.
It wasn't there. It had never been there. Nobody had ever submitted it, and Bing's crawler — less aggressive than Google's, particularly with smaller sites in smaller markets — had never picked it up on its own.
That was the whole mystery. No penalty, no technical fault, no cloaking accusation. The site was simply never introduced to the second half of the searchable web, and nobody noticed for years, because nobody was looking. Every rank tracker, every SEO report, every dashboard we'd ever used measured exactly one thing: Google.
Why this matters more in 2026 than it did in 2020
If this were just about DuckDuckGo's single-digit browser market share, it would be a curiosity. It isn't, because the consumers of the Bing index are no longer primarily humans.
They're tools. Local LLM stacks defaulting to DuckDuckGo because it's keyless. Agent frameworks and wrappers shipping with the cheapest search provider wired in. Assistants and pipelines that hit Bing's API or its derivatives because access is straightforward and the pricing is sane. The AI layer of the web is being assembled, right now, largely out of whatever indexes are convenient — and convenience points at Bing far more often than the browser market share suggests.
Which produces the failure mode we walked into: a site can be perfectly optimized, structurally sound, schema-complete, fast, agent-ready by every on-page measure — and still be a null result for a meaningful slice of AI-mediated queries, because it's missing from the index those systems actually consult.
Google-only visibility measurement made sense when Google was the only consumer that mattered. That assumption expired quietly, and most measurement practice hasn't caught up.
To be precise about the claim: this is not "Bing has replaced Google," and not "every AI agent runs on Bing." Nobody has measured what share of AI-mediated retrieval actually flows through which index, and we won't pretend to know. The claim is simpler: the common assumption that visible on Google equals visible on the web is increasingly false. Google remains dominant, but it is no longer the only entry point to content — a fraction of engines, AI tools, agents, and local integrations draw on other indexes and data sources. Bing happens to be the example worth starting with, because it's easy to check, ships its own webmaster tooling, and quietly feeds visibility in places site owners don't expect. This piece isn't an argument for abandoning Google. It's an argument for no longer treating Google as the whole internet.
The fix (twenty minutes, no budget)
1. Check whether you have the problem. Search site:yourdomain.com on Bing and on DuckDuckGo. Empty result = you have the problem. Thirty seconds.
2. Bing Webmaster Tools. Microsoft's counterpart to Google Search Console. It offers a direct import from Search Console — site verification and sitemaps carry over automatically. Five minutes.
3. IndexNow. An open protocol through which your site pushes new and updated URLs to participating engines (Bing, Yandex, Seznam, Naver, among others) instead of waiting for a crawl. Several CMS plugins and CDNs support it natively; on WordPress it's typically a single toggle in an SEO plugin. If you run anything custom, the protocol itself is a trivial HTTP ping plus a key file at your domain root.
4. Check the door. Audit robots.txt and any WAF/CDN bot rules for bingbot blocks. "Block everything except Googlebot" configurations from the security-template era are a recurring cause of exactly this situation — and they make steps 2 and 3 useless.
5. Verify in a few weeks. Re-run the site: check. Our site appeared in Bing's index within weeks of submission; the same local assistant that had denied our existence then answered correctly.
The broader point
The lesson we took isn't "submit your site to Bing," although you should. It's this: visibility is now a property of indexes, plural — and of the growing set of machines that consume them. Any measurement practice that watches a single index is measuring the past.
We've folded multi-index presence checks into our own standing procedures, and we're exploring what continuous monitoring across engines should look like. If you're working on the same problem, we'd like to hear from you.
Senteri — how machines read the web.
