If you're trying to influence what ChatGPT says about a client's business, the first question isn't "what should we optimise", it's "where is it actually looking". We logged every source cited across thousands of AI responses to find out, and the answer is more structured than the "AI just scrapes the whole internet" assumption most people work from.
Two layers do most of the work
For our AI Visibility Report, we audited 10,000 US local businesses through the Insites 360-degree platform and recorded every source ChatGPT cited when answering awareness, reputation and recommendation queries. Across all of it, two source types consistently sat at the centre: the business's own website, and directory or listings platforms. Everything else played a supporting role.
The website is the anchor
It was the single most-cited source by a significant margin, appearing in 72.3% of awareness and reputation queries and still in 50.7% of recommendation queries. When AI is building a picture of what a business does, where it operates, and how it describes itself, it goes to the website first. This is also the one part of the equation a business fully controls, and the layer most likely to carry the accuracy errors we found in half of AI's answers.
Directories are the confidence layer
Directory and listing platforms appeared in 35-39% of queries, the second most common source type. Their value isn't really about frequency, it's about corroboration. A website can say anything about a business. When the same name, address and category show up consistently across multiple directories too, AI gains confidence that the information is actually correct. Inconsistent or fragmented listings do the opposite: they erode that confidence and can lead to weaker interpretation or exclusion from recommendations altogether.
The full citation breakdown
| Citation type | Awareness / reputation queries | Recommendation queries |
|---|---|---|
| Own website | 72.3% | 50.7% |
| Directory / listings | 38.85% | 35.25% |
| Travel sites | 17.63% | 12.95% |
| Trades sites | 10.07% | 23.02% |
| Review sites | 1.8% | 11.51% |
| Professional sites | 1.8% | 3.96% |
| News sites | 0.72% | 3.6% |
| Social media | 0.72% | 2.16% |
| Government portals | 0% | 2.88% |
| Jobs sites | 0.36% | 2.16% |
| Shopping sites | 0.72% | 1.08% |
Notice the shift between the two columns. Reliance on the website drops in recommendation scenarios, and reliance on review sites and trade-specific sites rises sharply (review sites jump from 1.8% to 11.51%, trades sites from 10.07% to 23.02%). That's not the website becoming less important, it's a signal that comparison and validation sources carry more weight once AI is choosing between several businesses rather than just describing one.
The rest of the web fills in the gaps
Beyond the two anchor layers, AI draws on a broader, industry-dependent mix: review platforms, trade and travel sites, and editorial content like blogs and listicles, particularly in unbranded queries where it's comparing options. A meaningful share of these citations come from the "long tail" of the web, smaller local blogs, niche industry sites, aggregator-style pages, used less to identify a business and more to understand what "good" looks like within its category. These sources rarely act as the foundation on their own. They build on what's already established via the website and directories.
Why this matters more than any single tactic
The core finding isn't which platform gets cited most, it's how they're used together. AI doesn't rank one page in isolation, it builds confidence through agreement across sources. The website defines the business. Directories confirm it exists and is consistent. Everything else adds context and reinforcement. Only when those signals align does AI confidently surface and recommend a business.
Where to start an audit
Given the hierarchy, the order of priority is clear:
- Audit the website first. Is the business description, service list and location information accurate, current and unambiguous, since this is what AI reads before anything else
- Check directory consistency second. Same name, address, phone number and category across every listing, not just the major ones
- Then look at category-specific sources. Review platforms and trade or travel sites carry more weight in recommendation queries than in simple awareness ones, so their importance depends on the business type
Insites' AI visibility audits already surface this exact citation breakdown for every client, so agencies don't have to log AI responses by hand to see where a business actually stands. This hierarchy is also why agencies are having to rebuild their AI search offering around ongoing signal-building rather than one-off technical fixes.
This is why visibility in AI search feels different from classic SEO. It isn't about winning one ranking position, it's about creating a consistent, verifiable presence across the web, and the website is the cornerstone of that presence.
This analysis is drawn from Insites' AI Visibility Report, based on a clean audit of 10,000 US local businesses across ChatGPT and Perplexity. Run a free AI visibility check on any business in seconds.
Andrew is the co-founder and CEO of Insites, the AI visibility platform that agencies, telcos and media companies use to show local businesses how search engines and AI tools like ChatGPT and Gemini find, trust and recommend them. He has spent more than 15 years in digital marketing and UX, and co-hosts The Unusable Podcast on UX design.
