Capacity and engines
Help
·
Capacity and engines
·
Updated
The eight engines
Perplexity, Google AI Overviews, Google AI Mode, ChatGPT, Gemini, Claude, Grok and DeepSeek.
Two of them are Google and they are not the same
Google AI Overviews is the answer box above ordinary search results. Google AI Mode is a separate conversational tab where Google answers at length.
Asked the same question on the same day, the two give different answers and name different brands. Neither stands in for the other, which is why both are listed.
Which ones search
Perplexity and AI Overviews search by definition. ChatGPT, Gemini, Claude and Grok decide per question. DeepSeek cannot search the web at all.
Every answer records how many searches were behind it, and says so when there were none.
DeepSeek, stated plainly
DeepSeek’s connection offers no search and returns no citations, so a DeepSeek answer reflects what the model was trained on rather than what the web says this week. Their own app has a search toggle a person can use; the connection does not.
It is still a real signal about what the model absorbed. It is not the same signal as the other seven, and the product says so on every DeepSeek answer rather than leaving you to discover it.
Two that are missing
Copilot and Meta AI. Neither has a way in that returns the answer a real user sees, leaving only scraping and consumer accounts, and Keylight does neither.
How they are read
Five are read through a wholesale data supplier from the interfaces people actually use. Three are read through their own paid developer connections. The methodology page says which is which.
Did this answer it?