Prompt-based AI visibility samples what AI models say; log-based AI visibility counts what AI crawlers actually did on your site.
Prompt-based vs log-based AI visibility is the core split in AEO tools. Prompt-based tools like Profound, Peec AI, Otterly, and Scrunch ask AI models sample questions and count brand mentions, which is sampling. Log-based tools like citAEOtion read real server traffic and count actual AI crawler hits from GPTBot, ClaudeBot, and PerplexityBot. citAEOtion measures crawler visits at $34.99 per month.
Quick answer: The difference between prompt-based and log-based AI visibility is measurement method. Prompt-based tools run a list of sample questions against ChatGPT, Perplexity, Claude, and Google AI Overviews, then report how often your brand appears in the answers. Change the prompt, region, or model version and the result changes, so it is an estimate. Log-based tools read your server traffic directly. Every AI system must fetch a page before it can quote it, and that fetch is logged as a real request from a named bot. citAEOtion installs a lightweight collector on WordPress and reports which crawler hit which URL, when, how often, the HTTP status, and the crawler category: AI Training, AI Search, AI Assistant, or Data Scraper. It costs $34.99 per month, versus $99 to $300 per month for prompt-based tools.
What prompt-based visibility measures
Prompt-based tools treat AI answers as the thing to measure. They build a set of questions your customers might ask, send those questions to the models on a schedule, and record whether your brand or a competitor gets named. The output is a share-of-voice chart: you appear in 22 percent of answers this week, a rival in 31 percent.
This has real uses. It shows how a model talks about your category, and it can flag when a competitor starts dominating a topic. But three things limit it. First, it is a sample, so the number depends on which prompts were chosen. Second, models are non-deterministic, so the same prompt can return different answers on different runs. Third, it tells you nothing about whether the AI ever crawled your site. A model can mention you from stale training data while ignoring your latest pages entirely.
What log-based visibility measures
Log-based measurement starts from a simple fact: AI engines cannot cite a page they never fetched. When GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Amazonbot, or Bytespider requests one of your URLs, that request is real and recorded. citAEOtion reads that server traffic and turns it into a dashboard.
Instead of "you appeared in 22 percent of sampled answers," you get "GPTBot fetched 512 pages this week, ClaudeBot fetched 47, PerplexityBot hit your blog 9 times and never touched your product pages, and 3 percent of those requests returned a 404." That is a census of AI activity on your own property, not a poll. Anthropic documents its own crawlers and how site owners can allow or block each one in its crawler support docs, which confirms these bots identify themselves by name in exactly the traffic citAEOtion reads. See how the collector works on the how it works page.
Side by side
| Question | Prompt-based | Log-based (citAEOtion) |
|---|---|---|
| What it measures | Brand mentions in sampled AI answers | Real crawler requests in server traffic |
| Method | Sampling, non-deterministic | Census, deterministic |
| Tells you which bots crawled you | No | Yes, by name and category |
| Catches Amazonbot, Bytespider, scrapers | No | Yes |
| Typical price | $99 to $300+/mo | $34.99/mo |
Why the crawler data comes first
Citation is a two-step chain. An AI system has to crawl your page, then choose to quote it. Prompt tools only look at step two, and only through a sample. If your pages are never crawled, no amount of prompt tracking will explain why you are invisible. Log data shows the gap directly: if PerplexityBot never requests your service pages, that is the problem to fix, and you found it in minutes instead of guessing at prompt results. This is the foundation of AI crawler tracking.
This is also why log data is cheaper to act on. You are not tuning a prompt list or arguing about sampling error. You see a page with zero AI crawler hits, you check its internal links and robots rules, and you fix the actual cause. It fits into the wider method of how to measure AEO results.
Where each one fits
Prompt-based tracking is worth running when brand-mention share is a reporting requirement and you have budget for a $99 to $300 per month sampling tool. Log-based tracking is worth running for everyone, because it is the ground truth underneath every AI citation, and at $34.99 per month it is cheap enough to keep on permanently. Many teams start with citAEOtion, confirm which AI systems reach their content, and only add prompt sampling later. A side-by-side of the best AI visibility tools shows where each method wins. Compare plans on the pricing page, and agencies can send branded crawl reports through the agency plans.
Frequently Asked Questions
Is prompt-based or log-based AI visibility more accurate?
Log-based is a census of real crawler requests, so it is exact for the question of who crawled your site. Prompt-based is a sample of AI answers, so it estimates brand mentions and varies with prompt wording and model version.
Can prompt-based tools tell me which AI bots crawled my site?
No. Prompt-based tools only observe AI answers to sample questions. They do not read your server traffic, so they cannot report which crawlers, such as GPTBot or ClaudeBot, fetched your pages.
Does citAEOtion track AI answer mentions too?
citAEOtion measures real AI crawler activity in your server traffic, classified by bot and category. It answers whether AI systems reach and fetch your pages, which is the step that has to happen before any citation.
How much does log-based AI visibility cost?
citAEOtion is $34.99 per month for a single site, with multi-site and agency plans and up to 29 percent off annual billing. Prompt-based tools typically run $99 to $300 or more per month.
Why do AI crawlers show up in my server traffic?
AI systems must fetch a page before they can summarize or cite it. Each fetch is a request from a named bot such as GPTBot, ClaudeBot, or PerplexityBot, recorded in your server traffic, which is exactly what log-based tools read.
Want the numbers for your domain? Start a live demo or view pricing.