Metrics

Every figure in the Log Hero reports is defined here: what goes into the numerator, what goes into the denominator, and what a delta is measured against.

What a hit is

One hit is one request your integration recorded — one line in your server log, with the fields Anatomy of a log line walks through. Requests are stored already grouped by day, so every figure is a sum of hit counts rather than a count of rows. Two consequences worth knowing:

• The finest granularity is a day. There is no hourly view, and there cannot be one from this data.
• Today is excluded. Today’s data is still arriving, so every chart and every figure ends yesterday.

Rules that apply everywhere

• Log Hero’s own crawler is excluded. KeywordHeroBot is filtered out of every figure, human-side ones included.
• Deltas compare like with like. A delta is always against the immediately preceding period of the same length. Seven days selected means the seven days before that, labelled vs previous 7 days.
• A hostname filter, once there is more than one. The selector appears in the top bar as soon as Log Hero has seen a second hostname.
• Assets are included unless a metric says otherwise. Images, CSS and JavaScript are requests like any other.
• An unnamed bot is still a bot. Requests whose User-Agent Log Hero could not resolve to a known name are grouped as Unknown bot and counted.

Bot Activity

Bot Activity. Note the units in the deltas: percent for hits and response time, percentage points for the error rate, a plain count for active bots.

Metric	How it is calculated	Its delta
Bot hits	Sum of bot hits in the window	Percent change
Error rate	Hits with a status of 400–599, divided by all bot hits. 3xx is not an error	Percentage points
Active bots	Distinct bot names with at least one hit. No minimum	Absolute difference
Avg response	Total response time divided by total requests measured — a hits-weighted mean, not an average of daily averages	Percent change

The four tiles follow the selected bot, but not the selected status class: filtering the page to 4xx does not turn Bot hits into a count of 4xx hits.

Server errors are in this number

The error rate covers 400–599, 5xx included. How completely 5xx reaches Log Hero depends on your integration: an edge or web-server integration records every one, while the WordPress plugin and the PHP SDK miss the errors a request never reached PHP for — a dead PHP-FPM pool, a gateway timeout. See Status codes for the breakdown per integration.

AI Visibility

An AI bot here means a bot categorised as AI training or AI assistant. Googlebot and bingbot are search bots and are not counted as AI.

AI Visibility. The fourth tile only appears once a sitemap inventory exists.

Metric	How it is calculated	Its delta
AI share of bot hits	AI-bot hits divided by all bot hits	Percentage points
User-triggered AI fetches	Hits from bots in the user-fetcher role — ChatGPT-User, Perplexity-User and the like	Absolute difference
AI bots hitting 404	AI-bot hits with status 404 exactly, divided by all AI-bot hits. Other 4xx are not counted here	Percentage points
Advertised pages AI has read	Sitemap URLs with at least one hit from an AI bot, divided by all sitemap URLs	None
AI avg response	As Avg response, over AI-bot rows only	Absolute, in ms

Crawl Budget & Waste

This report is per bot, and two of its figures come from a sample: the 5000 most-requested URLs in the window, not every URL. On a site with fewer than 5000 crawled paths the sample is the whole thing.

Crawl Budget & Waste. The hints under the tiles say which figures are sampled: “top 5 000 URLs”.

Metric	How it is calculated	Its delta
Useful crawl	Share of sampled hits that are 200-class and not an asset and carry no query string. robots.txt and sitemap requests are excluded from both sides	Percentage points
Crawl hits	Sum of hits for the selected bot. Exact, not sampled	Percent change
Sitemap coverage	Sitemap URLs this bot has hit, divided by all sitemap URLs	None
Distinct paths crawled	Distinct paths in the sample, so it stops at 5000	Absolute difference
Crawls per path	Sampled hits divided by sampled distinct paths — both sides from the sample, so the ratio stays comparable across site sizes	Absolute difference

Status wins over path type when classifying waste: a 404 on an image counts as a not-found, not as an asset.

Bots vs Humans

All three tiles count paths, not hits, and none of them has a delta. The hit figures underneath them are context.

Bots vs Humans. The small print under each number is the hit count behind the path count.

Metric	How it is calculated
Crawled and visited	Paths with at least one hit from the selected bot and at least one human hit
Crawled, never visited	Paths with bot hits and zero human hits
Visited, never crawled	Paths with human hits and zero hits from the selected bot

“Human” means “not identified as a bot”

It is a residual category. Unrecognised crawlers, login probes and prefetch requests all land in it, and human hits include assets and every status code. The day-by-day chart is the one exception: it counts only 200-class, non-asset requests, so its numbers are legitimately lower than the tiles’.

Coverage

Needs a sitemap inventory. Every figure compares the stored snapshot of your sitemap against hits in the selected window. Paths are normalised before matching — hostname lower-cased with a leading www. removed, query string and trailing slash dropped — so /page?utm_source=x and /page/ are the same URL here.

Metric	How it is calculated
Crawl Ratio	Sitemap URLs with at least one hit from the selected bot, divided by all sitemap URLs
Reached by any AI bot	The same ratio for the union of AI bots
Advertised, not crawled	Sitemap URLs minus those the selected bot has hit
Advertised, never requested by anyone	Sitemap URLs with no hits at all, from any bot or human

Coverage. Here AI bots have reached more of the sitemap than Googlebot has.

A ratio is always a union across the selected bots, never a sum, so it cannot exceed 100%.

What Changed

No tiles. The page compares the last complete seven days against the seven before them — a fixed window that ignores the date picker — and raises an alert when a movement clears one of these thresholds.

Alert	Threshold
A bot crawls more or less	±50%, and at least 100 hits in the larger week
A bot collapses	−90% or worse
A bot surges	+1000% or more
A bot appears or disappears	At least 50 hits on the side where it exists
Status mix shifts	1 percentage point of share, in either direction
Response time moves	±30%, with at least 200 hits in both weeks overall, or 100 per bot
A section starts 404ing	At least 20 404 hits last week, and a rise of 50% or more

Two weeks is a short baseline, and these thresholds are the same for every project. Read an alert as something to look at, not as a finding.

URL Drilldown

Metric	How it is calculated
Bot hits / Human hits	Hits on this exact path, split by whether the request was classified as a bot
Dominant status	The status code with the most hits on this path in the window
Avg response	As Avg response, for this path only

Exact match, so query-string variants are separate URLs and parameter URLs cannot be drilled into yet — their hits count against the base path.

URL Drilldown. “28,861 of 28,861 responses” under the dominant status means every single request got the same code.

What none of these figures can tell you

• Errors that never reached your application. On the WordPress plugin and the PHP SDK only. See the note under Bot Activity.
• Soft 404s. A page that returns 200 with “not found” content is a 200 here.
• Whether an AI assistant cited you. Logs record the fetch, not the answer.
• Whether a crawled page is indexed. Pair Coverage with Search Console.
• Anything below a day. No hourly patterns, no request-level timing.

Related

• Status codes
• The Log Hero reports
• Bot catalogue