Supported AI models
Scout queries the official APIs of three AI engines on every scan. This page lists exactly which model versions are in use today, and how that list will evolve.
The active model lineup
Each engine gets its own dedicated dashboard page so you can dig into its mention rate, sentiment skew and top-performing prompts in isolation:
ChatGPT
Gemini
Perplexity
| Engine | Model | What it represents | Why this version |
|---|---|---|---|
| ChatGPT | gpt-4o |
The model behind most consumer ChatGPT answers. | Highest-fidelity stand-in for what real users see. |
| Gemini | gemini-2.5-flash |
Google's flagship fast model — what powers Gemini answers and Search Generative Experience. | Switched from gemini-1.5-flash after Google retired the 1.5 family in September 2025. |
| Perplexity | sonar |
Perplexity's search-grounded model. Queries the live web for every answer. | Closest behaviour to live retrieval — best for citation tracking. |
Internal models (not user-facing)
Two more models run inside Scout, but they don't generate the answers you see — they help us extract structured data and write recommendations:
| Model | Used for |
|---|---|
gpt-4o-mini | Strict-JSON analysis of every AI response (mention detection, sentiment, position, brand extraction). |
claude-haiku-4-5 | The AI Audit Agent and LinkedIn content generator. Premium-only. |
We measure the engines that real buyers use. As of July 2026, ChatGPT, Gemini, and Perplexity own almost the entire generative search market for English-language buying questions. We'll add Claude.ai search, Grok, and AI Overviews when their query volumes justify the cost — and we'll always announce additions in the changelog before they go live.
Model upgrades and deprecations
AI providers update models constantly. Scout's policy is:
- If the provider releases a new major version of a model we use, we run it in shadow mode for 7 days, compare results, and switch over if quality is at-or-better than the old version.
- If the provider deprecates a model we use, we switch immediately to the official replacement and publish a note in your account's changelog.
- Historical scans are never re-run on new models — the data you saw on the day you scanned is the data we keep, so trend lines stay honest.
Claude (web search), Grok (real-time X/Twitter context), and Google AI Overviews are on our roadmap. Drop us a line at [email protected] if there's a model you need us to add for your industry.
Was this page helpful?
Thanks for the feedback — we read every response.