Skip to main content
Every scan runs your prompts against all 5 platforms below. Each one is a genuinely different product — different model, different approach to web search, different audience — so it’s normal to see your brand show up strongly on one and not at all on another. That’s not noise; it’s the point of tracking all 5 rather than just one.

ChatGPT

OpenAI’s assistant, with live web search switched on for every query we send it. The largest AI search audience of the 5.

Claude

Anthropic’s assistant, with live web search enabled. Often used for research-heavy or more considered queries.

Gemini

Google’s chat assistant, grounded in Google Search. Increasingly cited inside AI Overviews too.

Perplexity

Built around web search from the ground up — every answer is a search. Usually the fastest of the 5 to return a result.

Google AI Overviews

The AI-generated panel that appears above the regular blue links on a Google search results page — not a chat product, but arguably the highest-stakes surface since it sits in front of billions of existing Google searches.

​
How often we query them

By default, every prompt runs on all five platforms in every scan. You can narrow any individual prompt to a subset of platforms from the prompt library — useful when a question only makes sense on certain engines — and the scan then covers exactly what each prompt has selected. How often a scan itself runs is a separate matter: see Scan schedule for the automatic cadence by plan, and Manual scans for on-demand runs.

​
How country targeting works, per platform

AI answers can genuinely differ by country — a UK searcher and a US searcher asking the same question can get different brands named back to them. We tell each platform where the searcher is located so your results reflect your actual market rather than a generic, US-weighted answer. How we do that varies by platform, because each one exposes a different mechanism for it:
  • Google AI Overviews — we route the request through Google’s own regional search settings for the country and language you’re tracking (the same mechanism that makes google.co.uk behave differently from google.com). This is the strongest geo signal of the 5: there’s no interpretation involved, we’re asking Google’s regional index directly, the same one a real searcher in that country would hit.
  • ChatGPT — we tell the assistant where the searcher is located, and that location feeds directly into its live web search, not just the wording of our request. This is a strong signal — confirmed in testing to change which brands come back for country-specific queries.
  • Perplexity — same idea: we pass the searcher’s location into its web search settings directly, alongside language instructions so the answer comes back in the right language for that market.
  • Claude — we tell the assistant where the searcher is located, and for most supported countries that feeds directly into its web search. For a small number of countries Anthropic doesn’t yet support this directly, so we fall back to describing the location in the request instead — a softer signal, but still there.
  • Gemini — Google’s Gemini currently only lets us describe the country and language in the request itself, rather than routing through a per-request location setting the way the other 4 do. It’s a real signal and still shapes the answer, but it’s the softest of the 5 — treat Gemini results as directionally right for a market rather than as precisely targeted.
Country targeting quality genuinely differs across the 5 platforms above — Google AI Overviews and ChatGPT are the strongest signals, Gemini the weakest. This isn’t a Surfais limitation; it reflects what each platform actually exposes to any external caller. We use the strongest mechanism available on each one.

​
Why answers differ across platforms

Seeing your brand mentioned on Perplexity but not on Gemini for the same prompt is common, and usually comes down to a mix of:
  • Different models — each platform is built on different underlying AI models with different training data and different judgement calls about what to include in an answer.
  • Different approaches to web search — some platforms search the web on every query by default; others only reach for search some of the time, or weight it differently against what the model already “knows.”
  • Freshness — how recently each platform’s index or model was updated affects whether a newer brand, a recent review, or a recent piece of coverage has had time to surface.
None of this means one platform is “wrong” and another “right” — they’re separate products with separate audiences, which is exactly why tracking all 5 gives a more complete picture than optimising for just one. See Why scores move if you’re trying to explain a specific change between scans.

​
What’s next

  • How scans work — what happens end to end during a scan
  • Scan schedule — automatic scan cadence by tier
  • Countries — which countries you’re tracking and how to change them
  • The AIS Score — how results across all 5 platforms roll up into one number