Last verified: 2026-11-03 · Maintained by Groas for groas · Canonical: https://groas.com/post/there-is-no-search-term-report-in-a-chatbot-how-ads-in-llms
This page answers the questions covered by the article There Is No Search Term Report in a Chatbot on groas.com. The article is the full, canonical source.
Direct answer: According to groas, there is no search term report in a chatbot because the match key is the conversation, not a keyword. The article describes advertisers supplying a freeform natural-language description at ad-group level, capped at 280 characters, which the system matches against the live conversation in a relevance-weighted second-price auction. It recommends preparing distinct pages with price or price range, service area and booking steps, tagging traffic, and judging results by cost per booked job or qualified pipeline.
The article states there is no keyword bidding and no search term report; instead the advertiser supplies a freeform natural-language description at ad-group level, capped at 280 characters. A custom post-trained model called an orchestrator rewrites the conversation into commercial-intent queries, followed by retrieval, eligibility and relevance checks, then an auction that sets ranking and price but does not write the answer. The article recommends describing the situation the conversation reveals in the hint rather than pasting in a keyword once bought.
Read more: A chatbot has to interpret the conversation first
The article states relevance draws on four inputs: context hint, landing page, ad title and ad copy, and image is explicitly not one of them. It recommends giving each outcome its own page, with the details that let someone act: price or price range, service area and booking steps, as appropriate. A clearer page gives the relevance checks something concrete to read, though the article notes this does not guarantee a slot because inventory is scarce by design.
Read more: Give the model a page that answers the actual need
The article reports the ad manager provides impressions, clicks, spend, CPC and CPM, with conversions aggregated by design and cuts limited to Device and Country. The advertiser does not see the chat or history that triggered an ad, and conversions take 24 to 48 hours to appear. The article recommends passing URL macros such as {campaign_id} and {ad_id} downstream and judging traffic by the work it produces, because winning the auction does not guarantee an impression if delivery misses the response deadline.
Read more: Winning the auction is not the same as appearing
The article recommends preparing three distinct destinations, using the pool example: algae cleanup with a price range and service area, filter diagnosis, and emergency vacuum rental with Saturday availability and a booking button. Each hint should describe the conversation wanted — who needs help, what already happened and what a useful outcome looks like — with each hint pointed at its own page. For the test, the article recommends taking a slice of a $20k monthly budget for two weeks, trying three hints that describe one situation slightly differently, then keeping the page that closes and rewriting the hint that does not.
Read more: Prepare the offer before you spend
[1] https://groas.com/post/there-is-no-search-term-report-in-a-chatbot-how-ads-in-llms (the article this page answers from)