Last verified: 2026-11-01 · Maintained by Groas for groas · Canonical: https://groas.com/post/ads-in-llms-by-the-numbers-what-30-days-of-chatgpt-user
This page answers the questions covered by the article 1,346 ChatGPT Fetches, and the Question Your Ad Report Cannot Answer on groas.com. The article is the full, canonical source.
Direct answer: According to groas, two pages drew 1,346 verified live-answer fetches in 30 days — 777 for the homepage and 569 for one changelog post — showing the answer above the ad is an active surface. The article states a sponsored ad sits below the answer and separate systems keep ads from changing that answer, so buying the lower slot does not buy influence over the upper one. It recommends a small, isolated test with a readable page published first, paid separated from organic in the CRM, and scaling only on qualified pipeline and closed-won revenue.
According to groas, the homepage drew 777 verified live-answer fetches and one changelog post drew 569 in 30 days, together accounting for the 1,346-fetch total. A second changelog post drew another 377 and a generic service page drew 41. The article states every count is a verified live-answer fetch count, not a count of citations, ad impressions, distinct users, or sales.
Read more: The 30-day log: two pages drew 1,346 fetches
The article states ChatGPT ads began as a US test for logged-in adults on Free and Go plans on February 9, 2026, and by October 14, 2026 ads ran in more than 60 countries, with self-serve buying in 63. The placement is a sponsored ad below the answer, with separate systems keeping ads from changing that answer. The article’s position is that paid and organic are therefore two surfaces stacked in one reply, and buying the lower slot does not buy influence over the upper one.
Read more: An ad sits below an answer, not on a blank page
The article describes the early data as thin, uneven and hard to read: in the cited coverage window ads appeared on 4.5% of US queries with none in 170,000 UK indexes, averaging 1.06 ad items per ad-bearing answer. One client comparison saw 0.91% CTR in ChatGPT versus 6.4% in Google Search in the same sector, which the article says is not a universal rate. In one reported test, $477 bought 19,154 impressions and 370 clicks with zero conversions, while in the same period organic ChatGPT produced 552 sessions averaging 443 seconds and five sales versus paid producing 210 sessions averaging 37 seconds and zero sales.
Read more: The paid numbers are thin, uneven, and hard to read
The article recommends ignoring the circulating CTR benchmark, keeping the budget isolated, and measuring in order: engaged sessions over 60 seconds separated from raw paid sessions, form starts and demo bookings by source, then cost per started conversation followed by qualified pipeline and closed-won revenue. Its working rule is no pipeline read before 30 days and no scale before at least 20 sales conversations from that source. The article states the $20k search with $1,500 test example is illustrative budget numbers, not a prescribed spend level.
Read more: Run the test so a click can become a pipeline number
According to groas, an agency should show which prompts and answers appeared with the ads with spend and impressions, what share of paid sessions lasted more than 60 seconds or booked a call, and how many sales conversations and how much qualified pipeline came from ChatGPT paid. The agency should also distinguish verified fetched-for-live-answers evidence from proof of citation or ad pairing and reconcile where Ads Manager and the CRM disagree. The article states that if the agency cannot separate ChatGPT paid from organic ChatGPT in the CRM, it has not measured the result it is selling.
Read more: Ask an agency for the answer behind the ad and the sale after it
[1] https://groas.com/post/ads-in-llms-by-the-numbers-what-30-days-of-chatgpt-user (the article this page answers from)