SemrushBot visits so Semrush can build its own SEO dataset (rank tracking, site audits, backlinks). It is an SEO tool crawler. Hardal's AI Visibility categories include SEO tools as their own bucket, distinct from trainers, searchers, and fetchers (What is AI Visibility?).
It does not buy ads and it does not carry oppref. Send its hits to OpenAI Ads, Meta CAPI, or GA4 as conversions and every cost-per-conversion you report comes out wrong.
How it shows up
User-agent contains SemrushBot (and related Semrush UAs). Volume can be large on competitive sites. It shows up in CDN and origin logs and nowhere else.
Treat it like AhrefsBot: infrastructure and crawl-budget, not acquisition.
What to report
- Request volume and bandwidth (this is a cost and WAF topic)
- Whether you allow it (teams that pay for Semrush usually do)
- That it is excluded from conversion destinations and from "AI visibility" narratives about ChatGPT citations
If SemrushBot is hammering staging or faceted URLs, that is a robots.txt / crawl-control ticket. If GPTBot is hammering /pricing, that is a training-corpus ticket. Those two tickets go to different people.
robots.txt
Allow if you want Semrush's own products to see you the way the tool expects:
User-agent: SemrushBot
Allow: /
Throttle or Disallow if the crawl is hurting origin and you do not need the tool's view of your site. This has no effect on Googlebot.
How Hardal measures it
Use the SEO-tools category in AI Visibility so SemrushBot is not plotted as OpenAI. Do not connect it to the OpenAI Ads destination. For actual ChatGPT paid media, how to measure ChatGPT Ads.