Compare · Promptwatch
Skald vs Promptwatch: both check the crawler, one keeps every request
Skald vs Promptwatch compares two tools that read your traffic logs and check the main AI crawlers against the addresses their owners publish. Promptwatch keeps the AI and search crawlers, counts people with a script, and reads AI answers in the real apps. Skald keeps every request in your log, counts people from it, and follows them to the lead.
Promptwatch is an AI visibility platform: it collects answers to your prompts from ChatGPT, Gemini, Perplexity, Copilot and Google's results pages, streams crawler logs from your CDN or server, counts AI-referred visitors with a script, and writes pages with its Content Agent. Skald reads your server's access log, sorts every request into five groups that add up to the log, and joins AI visits to your forms, bookings and payments. The other comparisons are listed on the Skald comparisons page. Every claim about Promptwatch below links to its own page, read on 26 October 2026.
30-day free trial · no card · we set it up within one working day · free to self-host
- crawler log stream · from your CDN
- visitor script · in your pages
- access.log · every request
ChatGPT-User GET /pricing/address checkedGPTBot GET /docs/address checkedPerplexity-User GET /guide/address checkedvisit /pricing/ from chatgpt.comChatGPTread /pricing/ 2 min beforevisit /guide/ no referrerdirectlikely Perplexityvisit /blog/ from google.comGooglevisit /docs/, no script rana person the script missed
3 visits, counted by the script · in these sample lines
Two records, kept in two places. The crawler stream holds the fetches, the script holds the visits by referrer, so the visit to /guide/ with no referrer stays direct.
Sample lines from one sample hour. Promptwatch's Page Tracker sets a page's citation crawls and citation clicks side by side; Skald's "likely" visits are never added to any figure.
Promptwatch docs, read 26 Oct 2026 · promptwatch.com/docs/academy/log-ingestion · promptwatch.com/docs/academy/visitor-analytics · promptwatch.com/docs/academy/attributing-ai-traffic
[01] question by question
Skald vs Promptwatch, question by question
Ten questions, each Promptwatch answer taken from its own page and linked. If its docs did not settle a question, the question is not here. How Skald checks a crawler's name is on AI crawler tracking.
| The question | Promptwatchby its own pages | Skaldyour server log |
|---|---|---|
| How it gets AI answers | For ChatGPT, "automated browser sessions that open the actual interface"; AI Overviews and AI Mode "from real Google Search results pages"; DeepSeek, Mistral and Claude through APIsPromptwatch docs, 26 Oct 2026 · promptwatch.com/docs/academy/how-promptwatch-collects-data | Asked through four assistants' APIs with your keys; each question 1 to 5 times, to see how much answers vary |
| Which AI models | Fourteen on its pricing page, from ChatGPT, Google AI Overviews, AI Mode, Copilot, Perplexity, Claude and Gemini to Llama, DeepSeek, Mistral, Amazon Alexa, Codex, OpenCode and Claude CodePromptwatch pricing, 26 Oct 2026 · promptwatch.com/pricing | Four: ChatGPT, Perplexity, Gemini and Claude. Not AI Overviews, AI Mode or Copilot |
| How often it asks | Daily: "The default, and what every plan uses, is daily"Promptwatch docs, 26 Oct 2026 · promptwatch.com/docs/academy/data-freshness | Set per tracking profile: daily, weekly or monthly |
| How it gets your site's traffic | Logs from Cloudflare, Akamai, AWS CloudFront, Fastly, Google Cloud CDN, Netlify or Vercel, a WordPress plugin, or an HTTP APIPromptwatch docs, 26 Oct 2026 · promptwatch.com/docs/academy/log-ingestion | Reads your web server's access log, a Cloudflare Worker's records, or a Vercel or Netlify log drain (the platform sends its request logs to Skald) |
| Which requests it keeps | Known AI crawlers and search engine bots; other requests "are dropped at ingestion, as are static assets"Promptwatch docs, 26 Oct 2026 · promptwatch.com/docs/academy/log-ingestion | Every request kept, people too, in five groups that sum to the log total |
| Is that crawler real? | "Promptwatch checks the client IP against those ranges"; unverified ones are left out of charts "unless you explicitly include them"Promptwatch docs, 26 Oct 2026 · promptwatch.com/docs/academy/log-ingestion | Nine owners checked: OpenAI, Anthropic, Perplexity, Google, Bing, Apple and DuckDuckGo by published list, Google, Bing, Apple and Yandex by reverse DNS, Meta by its network. A fake moves to bots as "claims to be GPTBot"; any other crawler keeps the name it sent, unchecked |
| How it counts people | "A lightweight script on your site reads the referrer of each visit"; AI referrer visits are "a floor"Promptwatch docs, 26 Oct 2026 · promptwatch.com/docs/academy/visitor-analytics | From the same log, with nothing in the page; AI clicks a floor too |
| After the click | "Promptwatch measures through the click; revenue lives in your own analytics or CRM"Promptwatch docs, 26 Oct 2026 · promptwatch.com/docs/academy/attributing-ai-traffic | Forms, bookings and payments joined to the visit they came from |
| Writing pages | "The content agent turns citation gaps into new pages and refreshes", and publishes to WordPress, Webflow or FramerPromptwatch, 26 Oct 2026 · promptwatch.com | Skald writes no content |
| What it costs | Essential $95 a month with 50 prompts, Professional $245, Business $579, Enterprise custom; Agent Analytics from Professional; a 7-day free trial. As listed on promptwatch.com on 26 Oct 2026Promptwatch pricing, 26 Oct 2026 · promptwatch.com/pricing | Free on a server you run; hosted Skald is charged per site, see the Skald pricing page |
Every Promptwatch answer was read on promptwatch.com and its docs on 26 October 2026; prices are its monthly brand plans in US dollars. If we have one wrong, tell us.
[02] crawl, cite, click, revenue
Four links from an AI crawl to revenue, and who runs each one
Promptwatch's own playbook lays out the chain: a crawler reads your page, a model cites it, a reader clicks, and some of them buy. It tracks the first three links itself. Pick a link to see where each tool gets it.
pick a link
- crawl
- cite
- click
- your CRM
- read
- cited
- visit
- lead
- Promptwatch, by its docs
- Crawler logs streamed from your CDN, its WordPress plugin or its HTTP API, addresses checked where the provider publishes its ranges.
- Skald
- Your access log: each AI fetch, its address checked for nine owners, with its status, size and answer time.
Promptwatch docs, Attributing AI traffic to revenue, read 26 Oct 2026 · promptwatch.com/docs/academy/attributing-ai-traffic
Skald counts the four links on their own roads and joins them by page and day, never adds them up. Its answer link is a sample, read as a direction over weeks.
[03] to be fair
Where Promptwatch does more than Skald
Promptwatch is built around AI answers and the pages that win them. Pick it for answers read in the real apps, far more models, a writer that publishes, and more ways to send logs.
Answers from the real apps
It opens ChatGPT's own interface the way a person would, and reads AI Overviews and AI Mode on real Google results pages. Skald's answers come from the APIs, which its own guide warns can differ from the app.
Far more models
Fourteen on its pricing page, with Copilot, Alexa, DeepSeek and coding assistants, plus city-level prompts and shopping results. Skald asks four assistants.
A writer that publishes
Its Content Agent turns citation gaps into new pages and refreshes and sends them to your CMS. Skald writes no content.
More ways to send logs
Seven CDNs, each with a guide, plus WordPress and an HTTP API. Skald reads a server log, a Cloudflare Worker, or a Vercel or Netlify log drain, where the platform sends its request logs to Skald; that needs Vercel Pro or Netlify Enterprise.
Promptwatch, read 26 Oct 2026 · promptwatch.com/docs/academy/how-promptwatch-collects-data · promptwatch.com/pricing · promptwatch.com · promptwatch.com/docs/academy/log-ingestion
[04] use both
Promptwatch for the answers, Skald for every request and the lead
Skald reads the log you already have, so it also shows the days before you set anything up, from the logs still on disk.
Answers
Promptwatch reads what many models say about you each day, and writes the pages to change it.
Every request
Skald counts people, AI assistants, search engines, bots and scanners, and your own traffic, adding up to the log.
The lead
Skald joins the form, booking or payment to the visit that came from an AI answer, so you see which answer brought the lead.
[05] choose by the question
Promptwatch, Skald, or both
Pick by whether you want to change what AI answers say, or to count every request and what it led to.
- "Write pages that AI may cite." Promptwatch.
- "Are we in Copilot, AI Overviews and Alexa answers?" Promptwatch.
- "Count people with no script on the page." Skald.
- "Which AI visit became a booking or a payment?" Skald.
What it does not do
- Only four assistants are asked, all by API, so Skald has no answers from Google AI Overviews, AI Mode, Copilot or Alexa.
- Skald writes no content and publishes nothing to your CMS.
- No shopping results or ads read from AI answers.
- Skald's WordPress plugin, Skald Leads, sends forms, not requests: on a host with no readable log, only the script tag runs, and it counts people, not crawlers.
keep reading
More on Skald vs Promptwatch
questions
Skald vs Promptwatch FAQ
Straight answers, including what Skald does not do.
What does Promptwatch do?
By its own pages, read on 26 October 2026: it tracks your prompts across AI models and collects the answers from the real apps (some models by API), streams crawler logs from your CDN or server and checks crawler addresses, counts visitors sent by AI with a script on your pages, and writes and publishes pages with its Content Agent.
Do both check that a GPTBot is real?
Yes. Promptwatch checks the client IP against the ranges a crawler's provider publishes, and leaves an unverified one out of your charts unless you include it. Skald checks a crawler's address for nine owners: against the list OpenAI, Anthropic, Perplexity, Google, Bing, Apple or DuckDuckGo publishes, by reverse DNS for Google, Bing, Apple and Yandex, and by Meta's own network for Meta. It files a fake under bots as "claims to be GPTBot". Crawlers from other owners are counted by name, unchecked.
Does Promptwatch keep requests from people and other bots?
Its log ingestion drops requests whose user agent is not a known AI crawler or search engine bot, and static files. People are counted by its visitor analytics script, which reads each visit's referrer. Skald keeps every request in the log and counts people from it, with nothing needed in the page.
Does Skald read Google AI Overviews or AI Mode?
No. Promptwatch reads AI Overviews and AI Mode on real Google results pages; Skald asks only ChatGPT, Perplexity, Gemini and Claude, through their APIs. On the visit side, Skald's log names an AI Mode click when the udm=50 mark is on its referrer or landing address, while an AI Overview click looks like an ordinary Google click.
Can I run Skald on my own server?
Yes. Promptwatch is a hosted service; Skald can run on your own server for free, under the Elastic License 2.0 (source-available). It then reads that server's log, or the logs a Cloudflare Worker or a Vercel or Netlify drain sends it.
Can I use both?
Yes. Promptwatch takes crawler logs from your CDN and a script on your pages; Skald reads your server's log. Each runs on its own, and the crawlers they both see can be compared week by week.
$ tail -f access.log | skald
See what AI does with your site. Free for 30 days.
No card. We set up hosted Skald for you within one working day, or you self-host it free. Every line of your log, sorted and added up.