AI bots are already on your website. GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot: they crawl your pages every single day to feed the answers ChatGPT, Perplexity, Gemini, and Google’s AI surfaces give to millions of users.
And here’s the uncomfortable part: you almost certainly can’t see any of it.
That’s why today I’m super excited to introduce Agent Analytics by OtterlyAI, full visibility into every AI agent and crawler visiting your website, which pages they reach, and how that activity connects to your visibility in AI answers.
It’s available right now on every Standard, Premium, and Enterprise plan. In fact, we silently rolled it out last week, and before we even announced it, and many marketing teams had already found it and connected their websites. That tells you something about how badly marketers want this data.
Why we built it: the ROI question
Soon after we launched AI search monitoring, we kept hearing the same question from marketing teams (usually from their CFOs): “What’s the ROI of all this GEO work?”
It’s a fair question. AI search is still a largely brand-driven channel. You track brand mentions and sentiment, you track citations, and those metrics matter. But brand metrics alone don’t answer the business question.
To answer it, you need to see the whole journey:
- Brand tracking: are AI engines mentioning and citing you?
- Traffic tracking: are agents crawling your site?
- Business impact tracking: are those agents and visitors signing up, converting, buying?
Agent Analytics is the bridge between step one and everything after it. It’s the first major milestone on our way from a pure AI search monitoring product to true GEO end-to-end tracking: from brand mention to crawl to visit to revenue.
Why your web analytics can’t see AI agents
Most web analytics tools (GA4 included) are JavaScript-based: you insert a small script, and it tracks users as they engage with your pages. The problem is twofold:
- Agents and crawlers don’t execute JavaScript. They request your pages directly from the server, so your traditional web analytics never sees them. Not undercounted. Completely invisible.
- Even human tracking is leaky. Ad blockers and declined cookie consents mean your JavaScript-based tools miss a meaningful share of real visitors, too.
The result: agent behavior is a total black box, and user behavior is at least a partial one. I hardly know a single marketing team that has a system in place today capable of capturing real agent behavior. Agent Analytics fixes that, not with another script, but by reading the one source of truth that sees every request: your server logs.
What is Agent Analytics?
Agent Analytics shows you every AI agent and crawler that visits your website, in real time, based on your actual server log data. Inside your brand report you get three views:
- Overview: total agent visits, pages visited, top engines, trends over time, and your most-crawled URLs with a per-engine breakdown
- Pages: a searchable list of every URL agents have visited, with visit counts and engine distribution
- Agents: a full catalog of every bot we’ve identified on your site, categorized and counted
To give you a taste from our own website: on otterly.ai, Claude’s crawlers really like our pricing page, while ChatGPT spends most of its time on our homepage. Different engines, different behavior, different intent. And until now, nobody could see any of it.
Not all bots are the same (and this matters)
One thing we learned quickly: raw visit counts are meaningless unless you know what kind of bot you’re looking at. Agent Analytics categorizes every agent into three groups:
- On-demand AI fetchers (like ChatGPT-User): a real person just asked an AI about something, and the agent is fetching your page right now to answer them. This is live demand.
- Search index crawlers (like OAI-SearchBot): building and refreshing the engine’s index. Once you’re indexed, they come back rarely. That’s normal.
- AI training / data scrapers: collecting content for model training.
This is why you should only ever compare volumes within a category, never across. On our own site, ChatGPT-User is by far the most active bot, while OAI-SearchBot barely shows up. That’s not a weakness, it’s exactly how an index bot is supposed to behave. An on-demand fetcher visiting 10,000 times a day and an index crawler visiting 12 times a week can both be great news.
Nobody in the industry has definitive answers yet on what every bot pattern means, including us. We keep running GEO experiments to find out. But that’s precisely why the time to start measuring is now: you can’t interpret trends you never captured. Start with a baseline.
What you can do with Agent Analytics alone
You don’t need a single prompt set up to get value from your log data. These use cases stand entirely on their own.
1. Find out which AI bots visit your site, and which don’t. The most basic question in AI visibility is the one most teams can’t answer: is my site being crawled by AI bots at all? Agent Analytics gives you a factual inventory: which bots show up, how often they return, and which parts of your site they care about. No guesswork, no third-party estimates. Just what your own server recorded.
2. Catch crawl problems before they cost you visibility. A bot that suddenly stops requesting your pages, or keeps hitting 404s on your most important content, is a visibility problem in the making. You just haven’t seen the downstream effect yet. Log data surfaces spikes, drops, and error patterns the moment they happen, so you can fix the cause weeks before it would ever show up as a missing citation.
3. Check whether bots actually respect your robots.txt. You’ve written your directives, but are AI crawlers following them? Your logs are the ground truth. Agent Analytics lets you verify compliance bot by bot. Paired with a crawlability check, you also see whether your robots.txt says what you think it says.
4. See how AI bots navigate your site architecture. Not all crawling is equal. Request patterns across your folder structure show which content clusters AI bots treat as important and which they barely touch. That’s a direct signal about how your architecture reads to a machine.
5. Start with a baseline, not a blank page. Connect a source, or upload historical log files, and you have data to work with before you’ve configured anything else. “Here’s exactly what AI bots did on our site last month” beats assumptions every time.
6. Measure the effect of technical changes immediately. Updated your sitemap? Fixed crawl errors? Bot behavior in your logs is the fastest feedback loop you have. You’ll see within days whether crawlers picked up the change, instead of waiting for downstream visibility data to move.
What you unlock by combining it with AI search tracking
This is where it gets really interesting, and where Agent Analytics becomes something no standalone log tool can be. Logs tell you what bots requested. AI search tracking tells you what AI engines cited. Together, they answer questions neither can answer alone.
Your citation report now shows enriched agent-crawl data for your own URLs, so you see both signals side by side.
1. Separate access from impact. Heavy crawling feels like success. But a crawl means a bot fetched your content, not that any AI answer used it. Combining both views turns two disconnected metrics into a funnel: crawled → cited.
2. Diagnose underperforming content with precision. When a heavily crawled page rarely shows up as a source, there are three very different explanations: a technical access problem, a content quality problem, or a gap in which prompts you’re tracking. Each has a completely different fix. With logs and prompt data side by side, you can tell them apart instead of guessing.
3. Pressure-test your prompt coverage. Prompt tracking has a built-in blind spot: it only reflects the topics you chose to monitor. Your logs don’t have that bias. Here’s a real example: our own pricing page is one of the most-crawled URLs on otterly.ai, but it barely appeared in our citation data. Why? We weren’t tracking the prompts people actually use before landing on a pricing page. The crawl data told us exactly where to expand our prompt set with prompt research. When bots pour attention into a content area your prompts don’t cover, that’s your cue, before you conclude the content “isn’t working.”
4. Make optimization decisions on evidence, not one-sided signals. High crawl volume with no citations looks like a win when it isn’t. Low citations without log context look like a content failure when the real problem is access. With both sides of the picture, you can prioritize the pages where effort will actually pay off. And when you ship a change, verify it moved the needle on crawling and citations.
Agent Analytics alone gives you observability: who’s crawling, what they can reach, whether your technical setup is helping or hurting. Add AI search tracking and you get attribution: whether all that bot attention converts into real presence in AI answers. One tells you the door is open. The other tells you people are walking through it.
Setup takes minutes
Because Agent Analytics works from server-side data, setup looks a little different from your typical analytics tool. Head to Data Sources inside your brand report and pick the option that fits your stack:
- Cloudflare Worker: forwards only AI-bot traffic; your normal site traffic is untouched
- WordPress plugin: install, activate, done
- Netlify Edge Function: a small copy-paste function
- Webhook: for any custom stack or provider
- Log file upload: upload historical logs (.csv, .json, .log, .gz, and more) for instant, retroactive analysis
Running on Webflow, HubSpot, or something else? Tell us. We’re prioritizing new integrations based on exactly this feedback. Full setup instructions are in our help center.
A note on privacy, because we know you’ll ask: Agent Analytics is fully GDPR compliant. We do not process any personally identifiable data, at all. Your users’ data never enters the picture.
Pricing and availability
Agent Analytics is available on Standard, Premium or Enterprise plans, starting today. No add-on, no extra product to buy. Each plan comes with a monthly event allowance; if you run a very large site with millions of requests, check the event limits on our pricing page. A Premium or custom Enterprise plan might be the right fit. Need help figuring it out? Book a call with us here.
What’s next
Agent Analytics is step one, deliberately. Here’s where we’re heading:
- Content intelligence (later this quarter): understanding why AI engines quote certain passages from certain pages, and helping you predict which content will be cited before you publish it
- Crawl-to-citation conversion rates: turning the crawled → cited funnel into a metric you can optimize against
- Deeper funnel tracking: connecting agent activity and AI user referral traffic all the way to signups and revenue
The end goal hasn’t changed: help marketing teams go from brand tracking to traffic tracking to business impact tracking, and finally answer the ROI question with data instead of hand-waving.
Start measuring today
Every AI visibility program needs a starting point, and right now, most teams don’t even have a baseline. AI agents are on your website today, shaping how millions of AI answers describe your brand, and your analytics can’t see them.
Change that in the next ten minutes: Try Agent Analytics for free
