[slau]

You can opt-in to get an LLM response by phrasing your queries as a question.

Searching for “who is Roger rabbit” gives me Wikipedia, IMDb and film site as results.

Searching for “who is Roger rabbit?” gives me a “quick answer” LLM-generated response: “Roger Rabbit is a fictional animated anthropomorphic rabbit who first appeared in Gary K. Wolf's 1981 novel…” followed by a different set of results. It seems the results are influenced by the sources/references the LLM generated.

[abtinf]

You don’t have to phrase it as a question; just append a ?, which is an operator telling it you want a generated answer.

[slau]

Yes. That is exactly what my answer demonstrates.

[abtinf]

Your answer is helpful, but "phrasing your queries as a question" might reasonably be interpreted as implying a full natural language question is required to trigger the LLM response -- especially since your example is a full sentence.

I'm trying to expand on your point by clarifying that "roger rabbit?" will also trigger the LLM response.

[pasc1878]

Even with a normal search there is a link to get Quick Answer which gives the LLM result

[greatgib]

I don't think that you are right. It is the search result that influence the llm generated result and not the opposite.

In your case, I think that it is just the interrogation point in itself at the end that somehow has an impact on the results you see.

[s900mhz]

It’s a feature of Kagi. Putting the question mark does invoke AI summaries.

https://help.kagi.com/kagi/ai/quick-answer.html

[greatgib]

I know for the summary.

What I say is that the search results part of the page, with or without the summary should be the same in theory.

So if the other person saw a difference in the result returned it might be only because of the impact of the question mark character itself on the search index

[s900mhz]

Ahh thanks for the clarification, I misunderstood!