What are the best AI visibility tools for technical GEO audits? How we run them
How we run technical GEO audits for clients: bot access, rendering, errors, crawl-to-citation, and which AI visibility tools give us the evidence to fix them.
A technical GEO audit is the first thing we do on most engagements, and it often changes the whole plan. Teams come in asking for more content. We open the logs and find that half the important pages are being turned away at the door. Content comes second.
Here is how we run the audit, and the tools that make it possible.
What the audit has to answer
We work through six questions, in this order:
- Can the bots get in? We check robots.txt, firewall rules, CDN challenges and rate limits against the named AI crawlers: GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, PerplexityBot, Perplexity-User and Google-Agent, among others.
- Do they see the content? If the useful text appears only after JavaScript runs, a bot may receive an empty shell.
- What errors do they get? A page that fails for AI bots but loads for humans is invisible to us until logs show it.
- Which fetched pages get cited? Fetched and never cited is an editorial problem. Never fetched is an access problem.
- Is the site itself in good shape? Linking, content depth and crawlability.
- Do fixes move anything? Visibility and referred visits, over weeks.
Why logs come first
A robots.txt review tells you what you intended. Logs tell you what happened. We run audits on Promptwatch because its Agent Analytics gives us real-time, IP-verified AI crawler logs, so requests with spoofed user agents are filtered out. We see error tracking with failure explanations, can stack path filters to isolate a section of the site, and can export to CSV when a developer needs the raw list.
The most useful view is the crawl-to-citation path, which gives a citation rate per page. Sort by pages with many fetches and a low citation rate, and you have a shortlist of pages the bots like to read and models decline to quote. That list usually tells us what to rewrite.
We connect the logs through whatever the client already runs: Cloudflare (Logpush or an auto-deployed Worker), AWS CloudFront, Fastly, Vercel, Netlify, Akamai, Google Cloud CDN, a custom HTTP route or the WordPress plugin. Crawler logs start on Professional, so for clients on Essential we use the free Agentic Readiness Check and a conversation about upgrading.
The on-site half
For the site itself, the weekly Site Health audit in Promptwatch runs 21 checks with a health score, a what-changed view since the last run, and a weekly email. The Site Content and Site Linking reports and the crawlability checks give us the structural picture. 360 Insights then joins crawler logs, citation clicks and AI-referred visits per page, and Topic Insights shows where a page underdelivers for its topic. That join is the part that saves time, because we are not stitching exports together by hand.
Where other tools fit
We also use or recommend others for specific jobs, and none replaces the log layer:
- Cloudflare's AI Crawl Control is free on every Cloudflare plan and shows AI crawler requests, status codes and robots.txt violations. It only sees traffic that passes through Cloudflare and tells you nothing about citations.
- Foglift offers a free technical audit covering SEO, AI readiness, performance, security and accessibility, which is handy as a first pass.
- Agentic fix tools such as Seology, Okara and Encited ship fixes as pull requests or through MCP. Useful after the audit, but they are not evidence tools.
- Enterprise technical SEO platforms such as Botify suit very large sites that already run them.
What we hand the client
The deliverable is a prioritized list, not a score. The Action Board in Promptwatch helps here: its Discovery agent reviews the data weekly and suggests up to five evidence-backed actions per run, and we add custom actions with owners and comments. Developers get specifics (which paths, which bots, which errors). Editors get the pages with high fetches and low citation rates.
For agencies running several client sites, the self-serve agency plans include crawler logs and site audits from Kick-off, at $199 a month. They do not include the automated Content Agent, so plan content separately.
What an audit cannot do
It cannot make a model cite you. Fixing access removes obstacles, nothing more. We say that plainly, and we re-run the panel after fixes so the client sees what actually moved, which sometimes is less than hoped. If you are starting an audit this week, Promptwatch is where we would begin.