BINGBOT · 2026-11-17

bingbot user agent — the biggest bot your census misses

Bingbot filed 1,028 pageviews here in seven days — 5.1× Googlebot's 203 — and every named census missed it, because it walks in wearing Mozilla.

HUMANby @mrrobots4h5 min read

Bingbot's user agent is Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; bingbot/2.0; +http://www.bing.com/bingbot.htm) Chrome/116.0.1938.76 Safari/537.36. The name you grep for sits in the middle of the string, not the front — so a census that groups by the first token never sees it. On this site, Bingbot out-crawled every named bot in the week measured: 1,028 visits against Googlebot's 203. Count bingbot by substring, and the biggest bot walks out of the anonymous pile.

The count below is live, not recalled: one week of this site's traffic lake, measured the morning this ran. Every number in it names its basis — the tracker classifies crawlers by user agent, and the class layer got Bingbot right (seo_bot). The named layer is what blew.

Bingbot, real count1,028meta-externalagent949PetalBot748Applebot277Googlebot203PerplexityBot185OAI-SearchBot162Bingbot, in a first-token census0
The census ledger, one week, one site. Every bar is a first-token name — the dashed bar is where a name-keyed count files 1,028 Bingbot visits.

What is the bingbot user agent string?

The whole claim sentence is short: Bingbot's real user agent is a Chrome-shaped string with compatible; bingbot/2.0 in the middle, and the keyword is the only place its name appears. Microsoft's own crawler documentation lists the shape; here is the exact string this site caught, verbatim from the traffic lake:

Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; bingbot/2.0; +http://www.bing.com/bingbot.htm) Chrome/116.0.1938.76 Safari/537.36

verify: the string above is copied byte-for-byte from this week's measured spans, not from Microsoft's docs — docs describe the shape, the lake keeps the specimen.

Two things hide inside it. First, it starts with Mozilla/5.0 — the same opening as every browser and half the scraper fleet on the web. Second, the version that matters is buried after a semicolon, where a naive parser splitting on spaces and slashes reads the first token as the bot's name. A census that does that ends up with a bucket called Mozilla/5.0 holding Bingbot's 1,028 visits next to twelve thousand genuinely anonymous hits.

Why your bot census misses bingbot

The mechanism is one line: name-keyed counting works for bots that name themselves at the front, and Bingbot deliberately does not. PetalBot, GPTBot, ClaudeBot and PerplexityBot all announce themselves up front — which is why the census post this site ran on Huawei's PetalBot could name it cleanly: four robots.txt asks in 28 days, and it read everything else on faith. Bingbot's UA follows the classic browser-compatible pattern instead, the same pattern that let it ride through decades of user-agent sniffing unbothered.

BotFirst-token censusKeyword censusWhere the difference comes from
Bingbot0 (files as Mozilla/5.0)1,028UA opens Mozilla/5.0, keyword mid-string
meta-externalagent949949self-declared at the front
PetalBot748748self-declared
Applebot277277self-declared
Googlebot203203self-declared
PerplexityBot185185self-declared
OAI-SearchBot162162self-declared
ClaudeBot149149self-declared
GPTBot144144self-declared
Bytespider7575self-declared

Same week, same lake, two counting methods. The class layer (search-bot vs AI-bot vs other) got Bingbot right all along — its 1,028 visits sit inside the seo_bot class of 4,211, about one visit in four. Only the name column lied.

That is the uncomfortable part: the class layer was right and the name layer was blind, and the mistake is invisible in the output. A dashboard that says Mozilla/5.0: 12,000 looks like noise, not like the single largest measured crawler on the property. The party was going fine; nobody noticed the guest of honour came in a paper bag.

How to count bingbot on your own site

The fix costs one line anywhere you keep logs: match on the substring, never on the leading token.

verify: last tested 2026-11-17 — the census numbers above were re-measured in this site's traffic lake the same morning this ran; the grep shape is standard log hygiene and the reverse-DNS step cites Microsoft's documented verification method.

And the honest ledger, wound open: our own census posts had this same blind spot. The PetalBot piece named its league table correctly and still would have told you Googlebot was the biggest search bot here — off by five to one. The earlier crawlers-versus-search-crawlers census had it too. The fix rides forward in every count from now on; the old numbers stay printed exactly as they were measured, because a correction that quietly rewrites history is just a second kind of costume. One thing held, though: the week Bingbot came dressed as Mozilla, it still minded its manners — and the IndexNow pings this site sends were answered, on time, every time.