Peekabot. Your page, the way a machine reads it.
Enter a URL, pick any of 64 bots, and get a machine readability grade for that page: every image the bot has no description for, every hole in your heading outline, and every line it never receives, marked on a screenshot of your own page.
Measured in a real browser, not guessed from the HTML.
- The human pass. A real browser loads your page. JavaScript runs, styles apply, lazy content settles, and we measure where everything actually sits: every image and how much of the page it covers, every heading and its level, the size and weight of every block of text.
- The bot pass. We request the same page with the bot's published user-agent string. For raw-HTML readers like GPTBot and ClaudeBot no JavaScript runs, so what the server sent is all there is. For documented renderers like Googlebot we render fully, because that is what Googlebot does. We also evaluate your robots.txt against the bot's token.
- The read through. From those two views we build what a parser can build: the heading outline, the images it has a description for, how much of your text is navigation and footer, and how many kilobytes of HTML it downloaded per thousand words.
- The grade. Five categories, one letter, for this one page. Every point deducted is named in the findings below it, so you can check the arithmetic.
Fair questions.
What does the grade mean?
One page, five categories, one letter. F means the bot does not receive the page at all, or most of the text is missing from what it does receive. D means it gets the page and can read the words, but not the structure and descriptions it needs. A means a parser gets a clean outline, described images, and a lean document. Every point deducted is listed by name next to the grade, so you can check the arithmetic yourself.
Why is my image count different from what I expected?
We count images as they actually render, not as they appear in the HTML, so hidden and icon-sized images are excluded. An empty alt attribute is accepted as a decorative declaration only on small images, where a purely decorative image is plausible. Above that we count it as undescribed, because a full-width banner is not decorative. We weight by rendered area, so a hero counts far more than a logo.
Can you tell if there is text inside my images?
No. We do not read pixels and we will not pretend to. What we can tell you is which images the bot receives no description of at all. If those images carry words, those words do not exist for any crawler.
How does Peekabot know what a bot sees?
We fetch your page twice. Once as a normal browser with JavaScript running, and once with the selected bot's published user-agent string and no JavaScript - because most AI crawlers read raw HTML and never execute scripts. We also read your robots.txt and evaluate the bot's token against your URL exactly as RFC 9309 specifies. What you see is what actually came back.
Why do most AI bots miss JavaScript content?
Rendering JavaScript at crawl scale is expensive, and operators like OpenAI, Anthropic, and Perplexity don't claim to do it. Independent measurement (Vercel and MERJ's 2024 crawler study) confirmed these crawlers download script files but never execute them. Googlebot and Bingbot are the notable exceptions - both operators document full rendering, and Peekabot treats them accordingly.
Is this exactly what the real bot gets?
Almost. We send the bot's published user-agent string, but from our servers rather than the operator's IP ranges. Sites that verify crawlers by IP address - some firewalls do - may respond differently to the real bot. When a firewall blocks or challenges our fetch, we show you that response, because it's a strong signal the real bot hits friction too.
Is it really free?
Yes. Scan any page, any bot, no account. If you want to see which bots actually visit your WordPress site - live, every page, every day - that's what the Viz plugin does.
AI Readability monitors: Peekabot that never stops watching.
Inside Viz, this exact scan becomes a monitor: you pick a page and a bot, and Viz re-scans it on a schedule - weekly on the free plan, daily on paid. Each monitor keeps the letter grade, a 0-100 score, the five category scores (Access, Structure, Images, Order, Efficiency), and how much of your visible content actually survives in the bot's view, trended over 30 days.
A theme update that hides your headings from parsers, a plugin that moves your article body behind JavaScript, a firewall rule that starts challenging crawlers - each shows up as a falling grade the next morning, not as a mystery traffic dip a month later. Monitor scans are Viz's own and never count toward your traffic analytics. The free plan includes 1 monitor; Lite runs 10 and Pro 100, daily.
Peekabot shows a snapshot. Viz runs the whole funnel.
These bots are already on your WordPress site every day. Viz shows which ones, on which pages, what the visitors they send go on to buy, and gives the agents among them tools to act.