Beyond Google: Optimizing for Bing, Perplexity, and the Indexes Behind AI Search

This month we pulled Forge and Smith’s AI citation data from Bing Webmaster Tools. Over the past three months, Microsoft Copilot and Bing’s AI-generated answers cited our site 3,688 times across 66 pages.

For a search engine that most marketers treat as an afterthought, that’s a lot of visibility.

The reason is that your customers may never open Bing, but the AI tool that they ask often searches an index other than Google’s. Copilot answers from Bing. Perplexity runs its own index. ChatGPT pulls from several sources, depending on how it’s used.

If your site isn’t in those indexes, or your setup blocks the crawlers that build them, your chances of being cited drop sharply, no matter how good your content is. 

This guide covers where AI search actually gets its answers, what Perplexity SEO and Bing SEO involve in practice, and how to measure whether any of it is working.

Where do AI search tools get their answers?

When AI search tools pull answers from the web, they rely on indexes and crawlers that differ from one product to the next: Google’s for Gemini and AI Overviews, Bing’s for Copilot, Perplexity’s own for Perplexity, and a shifting mix for ChatGPT.

The ChatGPT row is based on third-party research, not anything that OpenAI has documented, which is why it merits further discussion.

ChatGPT is the Hard One to Pin Down

For about two years, the assumption in SEO circles was that ChatGPT searched Bing. Recent evidence says otherwise. 

In July, French search consultancy Resoneo captured 1,249 ChatGPT answers and read a field in ChatGPT’s server stream that identified the sources behind each result. On free accounts in instant mode, about 64% of results came from an index that OpenAI built itself and about 35% were from scraped Google results. Paid accounts in thinking mode leaned harder on Google, at roughly 75%.

Bing hasn’t vanished from the picture. A log study by Peec found that ChatGPT’s Deep Research mode runs all of its retrieval through Bing, with no Google fallback. OpenAI says very little publicly, beyond its crawler documentation stating that sites blocking OAI-SearchBot won’t appear in ChatGPT search answers.

Treat those percentages as a guideline rather than a rule, since they come from outside researchers reverse-engineering a product that changes constantly. But the lesson underneath them is important: a single AI tool can draw on three different indexes depending on the mode, so optimizing for just one of them leaves gaps that you can’t see.

Perplexity and the Independent Indexes

Perplexity crawls the web itself and answers from its own index, which is why Perplexity SEO starts with letting PerplexityBot in. Brave runs another independent index, which it says covers more than 40 billion pages, and sells access to it through an API that other AI products build on. 

Neither one depends on Google or Bing, so visibility in those engines doesn’t guarantee visibility in their indexes.

Why does Bing matter again?

Bing matters for AI search because it powers Microsoft Copilot directly, supplies the results behind Yahoo and most of DuckDuckGo, and now gives site owners more detailed AI citation data than any other search engine.

By the usual measure, Bing is small. It handled just 4.5% of searches worldwide in August 2026, compared to Google’s 91.1%. But that figure only counts people typing into a search box. It doesn’t count questions answered inside Copilot, or by tools grounded in Bing’s results.

What changed this year is the reporting. In February, Microsoft launched an AI Performance report in Bing Webmaster Tools showing how often your pages are cited in Copilot, Bing’s AI summaries, and select partner tools. In June it added four more views: Intents, Topics, Citation Share, and Compare. Citation Share is the exciting one. It shows what portion of the citations for a given query point to your site—the closest thing we have to a share-of-voice metric for AI answers.

Our own report gave us valuable new information. Our site holds 21% of the citations for “managed website maintenance with hosting,” which says far more than a traffic-by-source/medium count from GA4. 

We learned that 89% of our citations link to blog posts, led by practical guides like our post on what every website privacy policy needs, while service pages earn about 1.5% of citations. On our site, at least, AI tools overwhelmingly cite pages that answer specific questions well. Those tend to be blog posts, not service pages.

Can AI crawlers actually reach your site?

Many sites are harder for AI crawlers to reach than their owners realize, usually because a robots.txt rule, a firewall setting, or a JavaScript-heavy page is turning the crawlers away.

Training Bots and Search Bots are Different

Major AI companies run separate crawlers for different purposes, including model training and search. 

OpenAI’s documentation lists GPTBot for training and OAI-SearchBot for search, and you can allow one while blocking the other. Perplexity recommends allowing PerplexityBot if you want to appear in its results. Its second agent, Perplexity-User, fetches pages when someone asks a question and generally ignores robots.txt, since a person triggered the request.

The common mistake is a blanket “block AI” rule. It keeps your content out of training data, which may be what you wanted, and out of AI search results, which almost certainly wasn’t.

Your Firewall May Have Decided for You

Cloudflare changed its default on July 1, 2025 to block AI crawlers on new domains. If your site joined Cloudflare after that date, or someone accepted the recommended settings, check your AI crawler controls. Search crawlers may be turned away at the network edge before your robots.txt file is ever read.

Content Behind JavaScript May Be Invisible

Google renders JavaScript. Many AI crawlers don’t, and several major ones retrieve HTML without executing the JavaScript that builds the page. 

A study by Vercel and MERJ found that ChatGPT’s and Claude’s crawlers download JavaScript files without running them, while Gemini uses Googlebot’s rendering. Peec’s 2026 logs found the same for ChatGPT’s Deep Research, which had read pages as plain HTML and never clicked anything. Pricing, service details, or FAQs that load through a script or hide inside tabs may simply not exist as far as these tools are concerned.

Typical server-rendered WordPress sites pass this test without any changes. Sites built as single-page JavaScript apps should be tested.

7 steps to get indexed beyond Google

Getting into the indexes behind AI search takes a few hours of mostly one-time setup, and it has to happen before any content work can pay off.

What about DuckDuckGo and Yahoo?

DuckDuckGo and Yahoo mostly run on Bing’s index, so the Bing steps above already cover them. Yahoo states that all of its algorithmic results come from Bing. DuckDuckGo runs its own crawler for some features but sources most of its traditional links from Bing.

Local results work on DuckDuckGo too, despite its privacy focus. It uses Apple Maps for map and address searches and works with the approximate location that your browser sends, which it discards after the search. That’s one more reason to keep your Apple Business Connect listing accurate.

How do you measure AI search visibility?

You can measure AI search visibility with free tools that you probably already have access to: Bing’s AI Performance report for citations, Google Search Console’s new generative AI reports for impressions, and Google Analytics 4 for the visits that follow.

Bing AI Performance

Bing’s report goes the deepest. It shows total citations, the pages cited, and grounding queries, which are the searches that AI ran to find your content. Read those queries closely. 

Ours are full of phrases that nobody would type, like “legal liability privacy compliance risks website privacy audit,” which is the AI rewriting a customer’s question into its own search terms. Reading them shows you how the tool interprets your topic, and which of your pages it thinks answers it.

Not every citation means something, either. Our reCAPTCHA footer notice, which is boilerplate text that appears on many sites, drew 97 citations in three months.

Google Search Console

Google launched generative AI performance reports on June 3, 2026, covering AI Overviews, AI Mode, and Discover. They show impressions by page, country, and date, but no clicks, and they’re still rolling out to a subset of sites. If yours isn’t showing the report yet, that’s likely why.

Google Analytics 4

GA4 now sorts visits from AI assistants into their own channel. Several guides claim it misses Perplexity, so we checked our own property: since June 1, every Perplexity visit to our site was tagged as AI assistant traffic. It was a small sample, three sessions, but the channel clearly recognizes Perplexity.

The catch is historic data. Visits from before the channel existed keep their old labels, so our earlier Perplexity traffic still sits under referral and “(not set).” That makes year-over-year or before-and-after comparisons misleading unless you account for the change. 

What’s no longer worth your time?

Much of the familiar advice about alternative search engine SEO is out of date, and some newer AI search tactics rest on assumptions that have already changed.

You can stop using these tactics right now:

• Exact-match keywords for Bing. The old advice that Bing rewards exact keyword matches still circulates. Microsoft’s own guidance on AI answers points to the depth of your content, clear structure, supporting evidence, and freshness.
• Social signals for Yahoo. Yahoo’s results come from Bing’s algorithm, so there’s no separate Yahoo ranking system to chase.
• Building for a single AI pipeline. A year ago, “ChatGPT uses Bing” was the advice. The evidence above shows how quickly that changed, which is a good reason to cover every index rather than whichever one people are talking about.
• Citation totals as a scoreboard. A citation for boilerplate text counts the same as one for your best guide in most reports. Track which pages and queries earn citations to find the real story of how your site is performing. 

Wrapping up

AI search doesn’t run on one index, and it won’t any time soon. Copilot answers from Bing, Perplexity from its own crawl, Gemini from Google, and ChatGPT from whatever mix suits the question. Getting into all of them just takes some light setup work. 

Verify your site with Bing, let the search crawlers in at every layer, make sure your content lives in the HTML, and claim your listings.

After that, measure performance. Bing’s AI Performance report is the most detailed AI citation data available right now, and the grounding queries will teach you more than the totals.

Once the crawlers can reach you, the writing starts to count. Our playbook on writing for AI search covers how to structure pages so AI tools can lift and cite them, while our guide to generative engine optimization covers the broader strategy.

If you’d like a hand with any of it, our team can help you grow your audience across search and AI.