How AI Assistants Search the Web (And What It Means for Your Visibility)
Claude, ChatGPT, Gemini, and Perplexity all search the web differently - and they don't all use Google. Here's what actually drives AI search visibility and what you can do about it.
Most SEO thinking is still built around Google. Get Google to index your pages, rank well in Google, drive traffic from Google. That’s the playbook, and for most sites it’s still largely correct.
But something has shifted. A growing number of people are now starting their research in AI assistants rather than search engines. They ask Claude or ChatGPT to recommend a tool, summarise a topic, or explain their options - and the answer they get determines which websites they visit, or whether they visit any at all. AI-generated answers are increasingly the top of the funnel.
The complicating factor is that these AI assistants don’t all search the web the same way, and several of them don’t use Google at all. Understanding which assistant uses which index - and what that means practically - is genuinely useful if you’re thinking about how new users find your product or content.
The search backends each AI assistant actually uses
This is not well understood, even among people who work in SEO. Here’s what the research shows:
Claude uses Brave Search. When Claude.ai searches the web, it uses Brave Search as its backend. This was confirmed by TechCrunch in March 2025, when web search launched for Claude, and independently verified through result analysis by researchers at Profound, who found a very high overlap between Claude’s cited results and Brave’s top organic results. Claude does not use Bing. It does not use Google directly.
ChatGPT uses Bing as its foundation. ChatGPT’s web search capability draws from Bing’s index, reflecting the Microsoft-OpenAI partnership. That said, ChatGPT applies its own algorithm on top, which means Bing rankings don’t translate directly - the same Profound research found significantly lower alignment between ChatGPT results and Bing’s actual rankings than Claude shows with Brave. Being in Bing’s index is necessary but not sufficient for ChatGPT visibility.
Microsoft Copilot uses Bing. Unsurprisingly for Microsoft’s own product, Copilot draws on Bing directly.
Google Gemini uses Google Search. Also unsurprising. If you’re already indexable by Google, you’re covered here.
Perplexity uses its own crawler. Perplexity runs PerplexityBot, its own web crawler, and supplements this with Bing and Google APIs. It has its own index built from that crawl. Perplexity also crawls aggressively - it has been the subject of significant controversy around whether it respects robots.txt, with Cloudflare publishing research in 2025 alleging it uses undeclared stealth crawlers to bypass blocks.
A note on Brave and Google
Brave Search operates from its own independent index - it doesn’t rely on Google or Bing to power its results. However, Brave does offer an optional “Google fallback mixing” feature in the Brave browser, where users can allow Brave to anonymously check Google for the same query when Brave’s own results need more depth. Brave frames this as a temporary measure while their index matures, and their stated goal is to reduce dependency on it over time.
Crucially, this fallback is a browser-side, user-controlled option - it is not part of the Brave Search API. When Claude queries the Brave Search API, it’s querying Brave’s own index, not a Google-mixed result set. For AI visibility purposes, Brave’s independent index is what matters.
Why Brave has become so significant
A development that flew under the radar for most SEOs: Microsoft retired Bing’s public search API in August 2024. This left Brave as the primary independent commercially available search API - meaning developers building AI applications that need real-time web search have few alternatives.
In a February 2026 post on their blog, Brave stated that the Brave Search API now supplies most of the top-10 LLMs with real-time web search data, and describes itself as the only search API with its own independent index that is commercially available at scale. This reflects a structural shift: because Bing closed its public API, Brave has become the default infrastructure for AI web search in a way that most people haven’t noticed yet.
The implication: Brave’s index is more important for AI visibility than its direct user numbers (around 20-30 million monthly searches) would suggest. When Claude searches the web, it’s querying Brave. But Brave’s reach now extends to many other AI tools as well.
How this changes the SEO picture
Traditional SEO advice points everything at Google. But if a meaningful chunk of discovery now happens through AI assistants, and those assistants use a mix of Brave, Bing, and their own crawlers rather than Google alone, the picture gets more complicated.
A few things follow from this:
Bing indexing matters for ChatGPT and Copilot. This has been discussed in SEO circles since ChatGPT launched, and it remains true. If your content isn’t in Bing’s index, it won’t appear in ChatGPT searches. Bing Webmaster Tools is the equivalent of Google Search Console for this - it lets you submit sitemaps and see what’s indexed.
Brave indexing matters for Claude. This is less discussed. Brave has its own independent crawler, and unlike Google and Bing there are no full webmaster tools - no indexing reports, no crawl stats, no way to see what’s been indexed. There is however a URL submission tool where you can request individual pages be crawled. Brave’s broader indexing also happens through the Web Discovery Project - an opt-in feature in the Brave browser where users anonymously contribute data about pages they visit. The practical implication: you have less visibility and control compared to Bing, but you’re not entirely without options.
Perplexity is worth treating separately. Allow PerplexityBot in your robots.txt, ensure your important pages are in the initial HTML (PerplexityBot doesn’t render JavaScript), and use clean structured markup. Being cited in Perplexity answers is a meaningful traffic source for content-heavy sites.
IndexNow - what it actually helps with
IndexNow is a protocol that lets websites notify participating search engines immediately when content is published or updated, rather than waiting for crawlers to find it. It was launched by Microsoft and Yandex, and as of 2025 is supported by Bing, Yandex, Naver, Seznam, and Yep.
A common claim is that IndexNow helps with AI assistant visibility broadly. That’s partially true but needs qualifying. Brave does not support IndexNow. The protocol has no official Brave participation. So IndexNow submission helps with ChatGPT and Copilot (via Bing) but does not directly accelerate indexing in Brave or Claude.
Cloudflare has a Crawler Hints feature that uses the IndexNow protocol - if your site is on Cloudflare, this can be enabled relatively easily, and it handles Bing submissions automatically when you publish new content. Worth enabling, but its direct effect on Claude visibility is indirect at best.
For Brave specifically, the best you can do is ensure fast page loads, clean crawlable HTML without JavaScript-dependent content, a valid sitemap, and a robots.txt that doesn’t accidentally block Brave’s crawler. Standard technical SEO hygiene. There’s no automated submission shortcut.
What this means practically for a new site
When aidata.stream was indexed by Bing but not yet fully indexed by Google, ChatGPT users were already finding it through web search - because ChatGPT draws on Bing’s index, and Bing had picked up the site while Google was still catching up. The two channels are genuinely separate: Bing indexing affects ChatGPT visibility, while Brave indexing affects Claude visibility.
New sites often appear in both Bing and Brave before Google simply because Google’s indexing of new domains can take weeks or months, especially for sites without strong backlink profiles. Bing and Brave tend to be faster to pick up new content.
The practical conclusion: don’t treat Bing as an afterthought just because its direct search traffic is small. For ChatGPT visibility, Bing is the index that matters. For Claude visibility, it’s Brave. Ensuring your site is technically crawlable - fast, clean HTML, proper robots.txt, valid sitemap - is the foundation for both, and Bing Webmaster Tools gives you the submission and reporting tools that Brave currently lacks.
What you can actually do
A practical checklist based on what’s confirmed to work:
For Bing / ChatGPT / Copilot:
- Submit your sitemap to Bing Webmaster Tools
- Enable Cloudflare Crawler Hints if your site is on Cloudflare (handles IndexNow automatically)
- Or implement IndexNow directly if you’re not on Cloudflare
For Brave / Claude:
- Submit key URLs via search.brave.com/submit-url
- Ensure Brave’s crawler isn’t blocked in robots.txt
- Serve important content in initial HTML, not JavaScript-rendered
- Fast load times, clean markup, valid sitemap
For Perplexity:
- Allow PerplexityBot in robots.txt explicitly
- Same technical requirements as Brave: static HTML, fast loads
For Google Gemini:
- Standard Google SEO applies
For all of the above:
- Use schema markup - it helps all crawlers understand your content
- Earn legitimate backlinks - they’re a signal for all search systems
- Write content that directly answers real questions - AI assistants favour clear, citable answers over vague editorial content
A broader shift worth watching
The consolidation of AI search on Brave’s index - driven by the closure of Bing’s public API - is something the SEO industry has been slow to notice. As more AI tools integrate web search, and as Brave is often the only viable option for developers building those tools, Brave’s effective reach keeps growing beyond its direct user numbers.
This doesn’t mean abandoning Google SEO. It means understanding that the search landscape has more than one relevant index, and that a site that ranks in Google but isn’t crawlable by Brave or Bing has a gap in its AI discovery story.
You can see this starting to play out in analytics data. Traffic source reports that previously showed only organic, direct, and referral are beginning to include claude.ai, perplexity.ai, and similar sources as discrete referrers. Tracking these as their own channel - rather than having them disappear into direct or referral - gives you a more complete picture of where new visitors are actually coming from, and which AI assistants are sending them.
Ready to see where your traffic is really coming from? Try AI Data Stream free →
Related Posts
Keeping Your AI Analytics Costs Low
With BYOK pricing, you control your AI costs directly. Here are practical ways to keep them minimal - from picking the right model to managing conversation context.
How System Prompts Work - And Why They Matter for AI Analytics
The system prompt is the difference between an AI that hallucinates plausible nonsense and one that stays grounded in your actual data. Here's how they work, and what good analytics prompts look like.
Writing Better Questions for AI Analytics Tools
Vague questions get vague answers - and with BYOK AI tools, they cost more too. Here's how to ask your analytics AI questions it can actually answer.