Two AI tools, two completely different source lists
You type the same question into Perplexity and ChatGPT. You get two answers. They cite completely different websites. Neither of them cites you. Sound familiar?
This happens constantly, and it confuses a lot of business owners and marketers who are trying to understand how AI search actually works. The short answer is that these two tools are fundamentally different products with different architectures, different data pipelines, and different ideas about what "a good source" means. Understanding those differences is the starting point for doing anything useful about your visibility in either of them.
How Perplexity finds its sources
Perplexity is a live search engine first. When you ask it a question, it runs a real-time web search, retrieves a set of results, and then synthesises an answer from those pages. The sources you see listed at the side are the actual URLs it fetched and read during that process.
This means Perplexity's source selection is heavily influenced by:
- Current crawlability. If your site blocks Perplexity's crawler (PerplexityBot), you will not appear. Full stop.
- Query-time search signals. Perplexity uses search index data, likely a blend of its own crawl and third-party index providers, to decide which pages to pull into its answer pipeline. Strong traditional SEO signals, backlinks, topical authority, page speed, still matter here.
- Page readability at retrieval time. When Perplexity fetches a page, it needs to extract the relevant content quickly. Pages that are clean, well-structured, and marked up with schema make that extraction far more reliable.
- Freshness. Because Perplexity is live, newer content can surface quickly. A well-optimised post published this week can appear in Perplexity answers within days.
Perplexity also re-ranks its sources based on how well each one answers the specific question. It is not just picking the top ten Google results and reading them out. It is actively scoring relevance at answer time.
How ChatGPT selects its sources
ChatGPT's behaviour depends on which mode you are using, and this is where a lot of the confusion starts.
The base model without web browsing
When web browsing is turned off, ChatGPT is not fetching anything in real time. It is drawing entirely from its training data, which has a knowledge cutoff. It may confidently cite sources it "remembers" from training, but those citations can be outdated, incomplete, or occasionally hallucinated. There is no live retrieval happening at all.
ChatGPT with web search enabled
When the browsing tool is active (which is now the default in ChatGPT for most queries), the model uses Microsoft Bing's index to retrieve pages and then synthesises from them. This is a significant detail. ChatGPT is effectively a Bing-powered retrieval system layered on top of a large language model. Perplexity uses its own index plus third-party data. That single difference in data source explains a huge chunk of the variation you see in results.
Bing's index has its own ranking signals, its own crawl priorities, and its own view of which pages are authoritative on a given topic. If your site ranks well on Bing, you have a meaningfully better chance of appearing in ChatGPT's sourced answers. If you have historically focused only on Google, you may be well-indexed there but barely present in Bing, which flows directly into ChatGPT's answer pool.
For a deeper look at how ChatGPT's own crawl fits into this, see our post on whether ChatGPT uses Google's index or its own crawl.
The role of structured data in source selection
Both platforms benefit from schema markup, but in slightly different ways.
For Perplexity, schema helps at the extraction stage. When Perplexity fetches your page, structured data like Article, FAQPage, Product, or Organization schema tells the system exactly what kind of content it is dealing with, who created it, and what the key claims are. That reduces ambiguity and increases the chance that Perplexity accurately represents your content in its answer. Pages with no structure are harder to parse quickly and may contribute less to the synthesised answer, even if they are technically included in the source list.
For ChatGPT, schema matters at two points: during GPTBot's crawl (which feeds OpenAI's own data pipelines), and during Bing's indexing (which informs web search answers). Bing has long supported schema markup and uses it in its own rich result features. A well-marked-up page is more likely to be accurately indexed and to have its content understood correctly by Bing's systems, which in turn improves how it performs as a ChatGPT source.
If you want to understand how Perplexity specifically makes citation decisions, our detailed breakdown of how Perplexity decides which sources to cite goes much further into that process.
Why the same query produces divergent results
Let's be concrete about the mechanisms at play when two different source lists appear for the same question.
Different indexes, different rankings
Perplexity and ChatGPT are drawing from different pools of indexed content. Even if both index your page, they may rank it differently based on their own signals. A page that ranks 3rd for a query in Perplexity's retrieval layer might rank 15th in Bing's index, putting it outside ChatGPT's retrieval window.
Different answer formats drive different source needs
Perplexity often produces heavily cited, reference-style answers where individual claims are pinned to specific sources. ChatGPT with browsing tends to synthesise more fluidly, sometimes pulling from fewer sources for a more narrative response. The format each tool is optimising for shapes which type of content it tends to retrieve. Listicles, step-by-step guides, and FAQ-style pages tend to do particularly well in Perplexity. Longer, authoritative narrative content often performs better in ChatGPT.
Crawler access and recency
If your robots.txt blocks PerplexityBot but allows GPTBot, you will appear in ChatGPT and not in Perplexity, or vice versa. Many site owners have not audited their crawler access settings recently. It is worth checking. Similarly, Perplexity's live retrieval means it can include a page published yesterday, while ChatGPT's training data has a fixed cutoff for non-browsing responses.
Query interpretation differences
These models interpret the same natural language query differently. Perplexity might parse "best project management tool for freelancers" as primarily a comparison query and retrieve review-style content. ChatGPT might interpret it as an informational query and retrieve more explanatory content. The same words, different retrieval strategies.
What this means for your visibility strategy
The practical implication is that optimising for AI search visibility is not a single-channel exercise. You cannot just focus on ChatGPT and ignore Perplexity, or the other way around. Each platform has its own requirements.
Here is a practical checklist:
- Check your robots.txt. Make sure you are not inadvertently blocking PerplexityBot, GPTBot, or Bingbot. If you are blocking any of these, you are opting out of large sections of AI search.
- Optimise for Bing, not just Google. Submit your sitemap to Bing Webmaster Tools. Improve your Bing presence. This directly improves ChatGPT visibility.
- Implement schema markup that aids fast extraction.
Article,FAQPage,Organization, andProductschema all help both platforms understand and cite your content correctly. - Write content that answers specific questions clearly. Both platforms reward content that answers a question well over content that vaguely covers a topic. Tight, specific answers to real questions outperform long, meandering articles.
- Build topical depth. A site with ten well-structured posts on a tight topic is more likely to be treated as authoritative than a site with one post on each of a hundred topics.
At FlinnSchema, we run free AI visibility audits that check your site's performance across both Perplexity and ChatGPT, including crawler access, schema quality, and content structure. It is a useful starting point if you are not sure where your gaps are.
Should you try to appear in both?
Yes, always. These platforms are growing their user bases rapidly and they serve different kinds of users. Perplexity tends to attract more research-oriented, technically sophisticated users who actively want sources. ChatGPT's answer quality is improving week by week and its user base is enormous. Gemini is also developing its own retrieval behaviour. Appearing in just one of these is a partial strategy at best.
The good news is that the fundamentals overlap significantly. Clean, well-structured content with proper schema markup, solid technical foundations, and genuine topical authority will help you in all of them. The platform-specific differences, such as Bing optimisation for ChatGPT or PerplexityBot access, are relatively quick wins on top of those fundamentals.
Our approach to AI visibility is built around exactly this: a foundation that works across all the major AI search platforms, with targeted adjustments for each.
Frequently Asked Questions
Why does Perplexity cite more sources than ChatGPT?
Perplexity is designed as a research tool and explicitly surfaces its sources as part of the user experience. It typically retrieves and cites five to ten sources per query. ChatGPT with browsing often synthesises from fewer sources and presents a more unified answer, so fewer citations appear even if more pages were consulted internally.
If I rank well on Google, will I appear in ChatGPT and Perplexity?
Not automatically. ChatGPT's browsing uses Bing's index, not Google's, so a strong Google ranking does not guarantee visibility in ChatGPT. Perplexity uses its own index. You need to actively ensure your site is crawlable and well-represented across multiple indexes, not just Google.
Does schema markup help with both Perplexity and ChatGPT?
Yes. Schema markup helps both platforms extract and understand your content more accurately. For Perplexity, it improves how your page is parsed at retrieval time. For ChatGPT, it improves how Bing and GPTBot index and represent your content. The same markup implementation helps both.
How often do AI search source lists change?
Perplexity's source lists change with every query because it performs a live retrieval each time. Two identical questions asked an hour apart can produce different source lists if new content has been indexed or if the retrieval ranking shifts. ChatGPT's browsing results are also live when that feature is active, so they change too, though potentially at a slightly different cadence depending on Bing's index updates.