AI-readable text ratio check
See what percentage of your page is text AI can read.
Why text ratio matters for AI search
AI search fetches a page's HTML and reads the text inside it. What it judges isn't the design you see on screen, but the characters actually present in the code.
The catch is that on most web pages those characters are only a fraction of the full HTML. The rest is markup that draws the layout, scripts that handle behavior, and style code. The lower the text ratio, the fewer real sentences AI has to grab onto.
This tool checks how much data AI can actually read from your page.
Many AI crawlers don't run JavaScript
Browsers run JavaScript to fill in the screen, but many AI crawlers read the HTML they receive and stop there. If your page paints its body text in later with JavaScript, it can look full of text to a human while that text is absent from the HTML AI receives. This tool measures the raw HTML before JavaScript runs, so it's close to what AI actually gets.
Frequently asked questions
How are the thresholds set?
They are an editorial standard set by weekerp, not a standard published by Google or any AI vendor. We anchor on pages that actually earned AI citations and apply different thresholds by language and page type. For the specifics, see 'What's the reasoning behind the thresholds?' below.
What's the reasoning behind the thresholds?
A weekerp blog post (Korean, article) that actually drew Google impressions and GPT citations over three months measured about 15.6% by characters. We set the Korean-article threshold so that page sits just above 'Sufficient' (the sample is still small, so we'll adjust as data accumulates). We did not use folklore figures like 'text-to-code 25% or more' — their source is unclear and they date to the static-HTML era, so on today's web most normal content pages fall short of them. We also apply different thresholds by language and page type, because Korean and Japanese pack more information per character and landing/tool pages are normally markup-heavy.
Why do I have to pick a language and page type?
Because the same ratio means different things. Korean and Japanese pack more information per character, and landing/tool pages are normally markup-heavy. Applying an article threshold to a landing page produces false alarms. That's why, without a type selected, we show only the number and no verdict.
What counts as text that AI reads?
On top of the visible body text, we count structured data (JSON-LD) and head meta information (title, description, Open Graph, and so on). Images, scripts, and style code are excluded.
What does this ratio measure, and how is it calculated?
It's a supporting indicator of how much information AI can read. It's calculated as the character share of AI-readable text (body, JSON-LD, head meta) against the full original HTML.
Is a higher ratio always better?
No. Once you clear the threshold, that's enough — raising the ratio further isn't the goal. This is a supporting indicator that doesn't decide AI citability on its own. When the body is very short but the ratio is high, we actually lower the verdict.
Do you store the URL or the result?
This tool fetches the public HTML of the URL once to run the calculation, and it never asks for an email or sign-in to see the result.
Read why text ratio matters
How AI search reads text, and why this ratio becomes a problem — explained.