Skip to main content
You choose which artificial intelligence (AI) answer engines your workspace tracks. A prompt is a question we ask AI engines for you. You can compare the results in your dashboard. As of November 10, 2026, you can choose from 39 current tracking options. That includes 24 web options and 15 direct model versions. Search modes and model versions count separately, so this is not a count of distinct platforms.

Turn on the models you need

1

Open your workspace settings

Choose your workspace, then open Settings → Workspace → AI Models. Each workspace keeps its own model choices.
2

Choose your models

Turn on the models you want to track. Changes save automatically. For models that use extra answer runs, review and confirm the amount shown.
3

Check your results

Your next scheduled run uses your enabled models. Open Recent chats to read the captured answers.
All plans include access to current models, with no model-slot limit. Each check uses your monthly answer-run allowance. Some models use several answer runs per check.
New workspaces start with ChatGPT, ChatGPT Web Search, and Google’s AI Overview enabled. Newly added models stay off until you enable them.

Understand web and API answers

Web options collect answers from consumer AI products and search experiences. An application programming interface (API) lets software ask a model directly. Options with the API badge track those direct model answers. Web and API answers can differ, even when they come from the same provider. For example, DeepSeek Web Chat and DeepSeek V4.1 Flash are separate tracking options. Bing AI Mode is also separate from Bing Copilot. Answers can vary by region, product settings, and time. Not every answer includes a citation: a website the AI pointed to as its source.

Browse web tracking options

Google’s search options appear as AI Mode and AI Overview in the app. DuckDuckGo’s search answers and Duck.ai’s model choices are separate options.
Country support varies by engine. Doubao, Baidu Wenxin, and Baidu AI Search use their mainland China experiences.

Browse direct model options

These 15 current options carry the API badge. Their version names help you distinguish them from web products and older models. The model list puts popular assistant families first, then groups related options by provider. The popularity order uses public web-traffic data, rather than workspace usage. It reflects assistant-family popularity, not measured demand for each API version. The current order uses a16z’s October 2026 report, based on August 2026 web traffic.

Keep or turn off legacy models

The Legacy badge marks these older versions:
  • DeepSeek V3
  • Llama 4 Maverick
  • Qwen 3.7 Max
  • GPT-5 Search
You see a legacy model only when your current workspace already has it enabled. It stays available for tracking while enabled. Other workspaces do not see it as a new choice.
Turning a legacy model off hides it from that workspace’s settings and model pickers. You cannot turn it back on through those pickers. Your saved chats remain available.
Current replacements are separate choices. Enabling one does not turn off an older model for you.

Understand unavailable models

Consumer Grok is currently paused and marked Unavailable. You cannot enable it for new runs. Grok 4.7 API remains a separate, available tracking option. Retired options are removed from model pickers. This does not delete saved chats.