# Should You Buy Ads in LLMs? Run the 30-Day Log Test First Last verified: 2026-11-01 · Maintained by Groas for groas · Canonical: https://groas.com/post/before-you-buy-ads-in-an-llm-run-this-30-day-log-test-on This page answers the questions covered by the article [Before You Buy Ads in an LLM, Run This 30-Day Log Test](https://groas.com/post/before-you-buy-ads-in-an-llm-run-this-30-day-log-test-on) on groas.com. The article is the full, canonical source. **Direct answer:** According to groas, advertisers should not buy ads in LLMs based on someone else's proof, including the article's 1,777 verified fetches in 30 days. The article recommends running the same 30-day check on the site's own traffic to count verified live-answer fetches and commercial share. Its operating threshold is 100 or more verified live fetches from one assistant in 30 days, at least 20% to commercial pages, and ads that can actually be bought on that assistant. ## How can I tell if AI assistants are fetching pages I sell from? The article recommends counting live-answer fetches, not training or index crawls: a live fetch fires when a person asks something and the assistant checks the site, while a crawler only shows pages exist. According to groas, a user-agent string alone is not proof, and unverified claims should be left out of the verified count. The article reports attackers cycled through about 32 fake AI-bot identities and 6.3 million fake crawler requests in a single day. Read more: [Question: Are assistants fetching pages you sell from?](https://groas.com/post/before-you-buy-ads-in-an-llm-run-this-30-day-log-test-on#question-are-assistants-fetching-pages-you-sell-from) ## What log data do I need to keep for a 30-day assistant demand test? The article recommends pulling 30 days of requests from wherever traffic already lands and keeping the user agent, source IP, and full URL path, plus the date range and log source. It says to set up three buckets before counting: verified live-answer fetchers, verified crawlers on a separate tab, and unverified claims that are kept but never reported as verified. According to groas, if a site is spending $20k a month on search and cannot get those three request fields, it should skip the ad test until it can. Read more: [Setup: Keep the log, the source IP and the page path](https://groas.com/post/before-you-buy-ads-in-an-llm-run-this-30-day-log-test-on#setup-keep-the-log-the-source-ip-and-the-page-path) ## How should I tag URLs before starting the 30-day count? The article recommends tagging every URL before day one as Commercial, Educational, or Utility and keeping the rule stable for all 30 days. Commercial means pages built to start a transaction, such as pricing, product, services, demo, checkout, and signup; Educational means pages built to explain, such as blog posts, docs, and guides; Utility means infrastructure such as robots.txt, sitemaps, and favicon. According to groas, changing what counts as commercial halfway through can make a rise in commercial share an edit to the spreadsheet rather than a change in assistant activity. Read more: [Tag URLs before day one](https://groas.com/post/before-you-buy-ads-in-an-llm-run-this-30-day-log-test-on#tag-urls-before-day-one) ## How do I measure weekly fetches and commercial share? The article recommends making one row per assistant and one column per week for verified live-answer fetches, keeping crawlers alongside but not inside that count. Commercial share is verified fetches to tagged commercial URLs divided by all verified live-answer fetches for that assistant: 400 fetches with 40 to commercial pages is 10%, while 400 with 180 is 45%. According to groas, 60 verified fetches with 25 to a pricing page is more useful than 600 to a meme post, and assistant fetches should be kept beside organic clicks to the same commercial URLs, such as 50 pricing-page fetches beside 5,000 clicks. Read more: [Measure weekly fetches, then commercial share](https://groas.com/post/before-you-buy-ads-in-an-llm-run-this-30-day-log-test-on#measure-weekly-fetches-then-commercial-share) ## Should we buy ads in LLMs if our logs show assistant activity? According to groas, the operating threshold is 100 or more verified live fetches from one assistant in 30 days, at least 20% to commercial pages, and ads that can actually be bought on that assistant. The article states the threshold is a gate for a small test, not a claim that ads will work. It warns against three false positives: a spike in robots.txt hits, a large unverified total treated as verified, and one viral blog post pulling 300 fetches while pricing and product pages sit at zero. Read more: [Decision rule: Test an available ad channel, or wait](https://groas.com/post/before-you-buy-ads-in-an-llm-run-this-30-day-log-test-on#decision-rule-test-an-available-ad-channel-or-wait) ## How do I test an agency pitch for ads in LLMs? The article recommends asking the agency to run the protocol on the site's own logs and return the verified table, separating live fetchers from crawlers, verifying source IPs, and splitting commercial from utility pages. According to groas, any surviving proposal should answer Inventory, Targeting, Measurement, and Money in writing, including pixel installation, UTM macros, separate click-through and view-through values, and a CPA target tied to pipeline or sales. The article warns a daily budget is an average: a $100-per-day setting can spend $200 in a day, up to 2x in one day and 7x across seven days. Read more: [Use the table to test an agency’s pitch](https://groas.com/post/before-you-buy-ads-in-an-llm-run-this-30-day-log-test-on#use-the-table-to-test-an-agencys-pitch) ## Sources [1] https://groas.com/post/before-you-buy-ads-in-an-llm-run-this-30-day-log-test-on (the article this page answers from)