Question
How Bing Copilot Selects Sources Compared to Perplexity
Chat assistants 5 min read
Microsoft gives site owners more to work with. As of October 2026, it offers a citation report, published formatting guidance and a fast way to signal content changes. Perplexity’s crawler documentation covers access and says little about how sources are selected. The practical comparison is therefore uneven: for Copilot, you can work from Microsoft’s own statements and data; for Perplexity, you work from its crawler rules, outside experiments and independent audits.
How Copilot reaches content
Copilot and Bing are reported together. Microsoft’s February 2026 announcement of AI Performance in Bing Webmaster Tools describes it as showing when a site is cited “in AI-generated answers across Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations.” The same announcement notes that Bing respects content owner preferences expressed through robots.txt and other supported controls.
The Tow Center for Digital Journalism observed a practical consequence in its March 2025 audit of AI search tools. Copilot was the only tool with a public crawler that none of the 20 publishers in its dataset blocked, which the researchers attributed to Copilot likely using the same crawler as Bing search, so that blocking it would also mean leaving Bing. That makes a decision about Copilot crawling hard to separate from a decision about Bing Search access; it does not mean that visibility in Bing carries over to Copilot.
How Microsoft describes source selection
A detailed description comes from Microsoft’s advertising arm. Microsoft Advertising’s guidance, published in October 2025, says that in AI answers “ranking still happens,” but it is less about ordering entire pages and more about which pieces of content earn a place in the final answer. “Assistants like Copilot break content down, a process called parsing, into smaller, structured pieces that can be evaluated for authority and relevance. Those pieces are then assembled into answers, often drawing from multiple sources to create a single, coherent response.”
From that description, Microsoft recommends clear headings, lists, comparison tables and question-and-answer formatting, schema markup, and visible content. It warns against hiding key answers in tabs or expandable menus and against relying on PDFs for core information. Bing’s AI Performance announcement adds its own suggestions: clear headings, tables and FAQ sections; examples, data and cited sources; and regular updates.
How Perplexity reaches content
Perplexity runs its own crawler. Its crawler documentation states that “PerplexityBot is designed to surface and link websites in search results on Perplexity” and that it is not used to crawl content for AI foundation models. A second agent, Perplexity-User, fetches pages when a user’s request prompts it, and Perplexity says it “generally ignores robots.txt rules.” Changes to robots.txt may take up to 24 hours to apply.
Unlike Copilot, which Tow’s researchers believed shares Bing’s crawler, Perplexity can be blocked separately from any traditional search engine, because PerplexityBot is its own crawler. That makes the decision to appear in Perplexity independent of your Bing or Google strategy.
What is known about Perplexity’s selection
Perplexity’s crawler documentation does not describe how it ranks retrieved pages. Outside evidence fills part of the gap.
The GEO paper, presented at KDD 2024, tested content changes on Perplexity itself. Adding quotations from credible sources performed best on the authors’ word-count visibility measure, improving it by 22% over the baseline, while keyword stuffing performed 10% worse than the baseline. In the Tow Center audit, Perplexity answered 37% of its source-identification queries incorrectly, compared with more than 60% across all tools tested.
Copilot and Perplexity side by side
| Bing Copilot | Perplexity | |
|---|---|---|
| Crawler | Bing's crawler, as Tow's researchers inferred | PerplexityBot, its own crawler |
| Can be blocked separately from a search engine | Not practically, by Tow's inference | Yes |
| User-triggered fetching | Not described in the sources here | Perplexity-User, which generally ignores robots.txt |
| Publisher citation data | Yes: AI Performance in Bing Webmaster Tools | Not described in its crawler documentation |
| Published format guidance | Yes: headings, lists, tables, Q&A, schema, visible content | Not published; GEO experiments favor quotations and sourced statistics |
| Signal for content changes | IndexNow, which Bing describes as keeping AI answers fresh | Not described |
| Audit result (Tow, 2025) | Declined more questions than it answered | 37% of queries answered incorrectly |
Measuring Copilot citations
Bing’s AI Performance dashboard gives publishers concrete citation data. According to the announcement, it reports:
- Total Citations: how often your content is shown as a source in AI-generated answers during a period.
- Average Cited Pages: the average number of unique pages from your site cited per day.
- Grounding queries: the key phrases the AI used when retrieving content that was cited, shown as a sample.
- Page-level citation activity: citation counts by URL.
Microsoft cautions that these counts reflect “how often pages are cited, not page importance, ranking, or placement.” For Perplexity, measurement means repeated prompts and your own analytics.
Keeping content fresh for Copilot
Bing recommends IndexNow for AI freshness. Its announcement says IndexNow “helps keep information fresh across search and AI experiences by notifying participating search engines whenever content is added, updated, or removed.” The IndexNow FAQ adds that each participating engine makes its own decision about whether to index a submitted URL.
What to do
- For Copilot: make sure Bing can crawl and index your pages, set up Bing Webmaster Tools, review AI Performance data, enable IndexNow, and follow Microsoft’s formatting guidance where it fits your content.
- For Perplexity: decide whether to allow PerplexityBot, check that firewalls allow its published IP ranges, and write sourced, quotable content.
- For both: keep facts current and clearly stated, and check how each tool describes your brand over repeated prompts.
Start with Copilot if you can only do one, because Copilot appears to draw on Bing’s crawler, which most sites already allow, and Microsoft gives you guidance you can follow and a report that shows results. Treat Perplexity as a separate decision with its own crawler, judged through repeated checks rather than a dashboard.