# How to tell real AI assistant fetches from spoofed bot hits

Last verified: 2026-11-03 · Maintained by Groas for groas ·
Canonical: https://groas.com/post/three-quarters-of-our-ai-bot-traffic-couldn-t-be-verified

This page answers the questions covered by the article [Only 25% of Our Claimed AI Bot Hits Verified. The Page-Level Data Told a Different Story](https://groas.com/post/three-quarters-of-our-ai-bot-traffic-couldn-t-be-verified) on groas.com. The article is the full, canonical source.

**Direct answer:** According to groas, only 3,872 of 15,510 requests to its robots.txt that claimed to be AI bots verified, about 25%, so raw hit counts are mostly noise. The article recommends splitting claimed hits from verified hits and verified live-answer fetches, and not counting permission checks as content reads. In its 30-day pull, the homepage led with 821 live-answer fetches while a dated post verified at 853 of 872 claimed hits (97.9%) with 529 live-answer fetches, which the article uses to choose pages to inspect, not to claim citations.

## Why can't I use raw AI bot hit counts as proof an assistant read my page?

According to groas, only 3,872 of 15,510 requests to its robots.txt that claimed to be AI bots verified, about 25%. The article states a User-Agent can be typed into a script, so it is a claim, not an ID, and must be checked against operator-published ranges or forward-confirmed reverse DNS. The article concludes a hit count alone cannot answer whether an assistant fetched a page it might use to answer someone.

Read more: [the article](https://groas.com/post/three-quarters-of-our-ai-bot-traffic-couldn-t-be-verified)

## What is the difference between claimed hits, verified hits and live-answer fetches?

The article defines three tiers in order: claimed hits as requests with an AI-bot User-Agent, verified hits as claimed requests whose source checks out against the operator, and live-answer fetches as verified requests from user-triggered answer fetchers. According to groas, a verified background-crawler request and a verified user-triggered request are different kinds of events and should not be rolled together. The article cites independent production-site measurements where 49.5% of 16,548 claimed crawler requests verified as a warning that even a live-sounding User-Agent needs checking.

Read more: [Three counts that should never share one column](https://groas.com/post/three-quarters-of-our-ai-bot-traffic-couldn-t-be-verified#three-counts-that-should-never-share-one-column)

## Which pages looked busiest in the logs but did not hold up after verification?

According to groas, /robots.txt had 15,510 claimed hits with 3,872 verified for a 25.0% rate, which the article treats as permission checks, not content reads. Site assets and icons had high hundreds of claimed hits each with single digits to low tens verified, roughly 1-4%, and near zero live-answer fetches. By contrast the homepage had about 2,600+ claimed with 2,026 verified and 821 live-answer fetches, while a dated post had 872 claimed with 853 verified (97.9%) and 529 live-answer fetches.

Read more: [The paths that looked busiest were not the cleanest](https://groas.com/post/three-quarters-of-our-ai-bot-traffic-couldn-t-be-verified#the-paths-that-looked-busiest-were-not-the-cleanest)

## Does a verified fetch prove my page was cited by an assistant?

The article states a fetch is a lead for content work, not proof of a citation. In its 30-day pull the dated post had 529 live-answer fetches while service pages barely registered, which the article uses only to decide what to inspect and improve next. According to groas, the logs cannot show which prompt triggered a fetch or that the post was cited, and a private prompt cannot be reverse-engineered from a server request.

Read more: [A fetch is a lead for content work, not proof of a citation](https://groas.com/post/three-quarters-of-our-ai-bot-traffic-couldn-t-be-verified#a-fetch-is-a-lead-for-content-work-not-proof-of-a-citation)

## What should I ask for before paying for an AI-visibility report?

The article recommends not paying for AI visibility measured only in hits. According to groas, a buyer should ask for claimed hits, verified hits with the IP-file date, and verified live-answer fetches by path, and if a tool or agency cannot produce that split its busiest URLs may be robots.txt and icons. The article states it uses the homepage total of 821 live-answer fetches and the dated post total of 529 to choose which existing page to inspect, not to claim a citation it cannot see.

Read more: [Buy the split, not the hit count](https://groas.com/post/three-quarters-of-our-ai-bot-traffic-couldn-t-be-verified#buy-the-split-not-the-hit-count)

## Sources

[1] https://groas.com/post/three-quarters-of-our-ai-bot-traffic-couldn-t-be-verified (the article this page answers from)