One of the frustrating parts of content SEO is that two pages can cover the same topic with the same keywords and the same schema, and one will rank while the other doesn't. When that happens, the difference often comes down to something structural editors have been measuring for decades: voice. How long the sentences are. How much passive voice the writer leans on. How often they address the reader directly. How many hedges — "very," "really," "basically" — creep in.
The Voice & Tone Analyzer measures all of that on any page and, optionally, compares it against the pages that currently rank for your target query. It ran in your browser, no signup, no paid API, no word-count cap.
What It Measures
Paste a URL and the tool returns six things that matter for how a page reads:
Flesch Reading Ease and Flesch-Kincaid Grade. The two most widely used readability scores. An Ease score of 60 or higher means "understood by 13–15 year olds"; most pages that rank for commercial queries sit between 55 and 70. Academic prose ranges into the 30s, which is a tell.
Sentence length distribution. Mean length, standard deviation, and the percentage of sentences over 25 words or under 10. Flat sentence length — short and long looking the same — reads like a textbook. Healthy published prose tends to have standard deviation between 6 and 12.
Passive voice rate. Sentences with "to be + past participle" as a share of total sentences. The detector is naive but serviceable; above 20% is the threshold where the copy starts reading like a pharma insert.
Hedge density. Word counts for "very, really, basically, actually, honestly, literally, arguably" and a handful of cousins. These are the words editors strike first. Under 3 per 1,000 words is confident writing.
CTA density. Action verbs like "buy, sign up, subscribe, download, get started, call, book, schedule" per 1,000 words. Landing pages live on 3+ per 1K. Blog posts tend to run at 0.5 or lower.
Pronoun balance. Counts of "you / we / I / they." The mix tells you who the page is talking to.
How to Use the Competitor Comparison
The comparison is optional but it's where the tool earns its keep. Paste your URL, then paste up to ten competitor URLs — the ones currently ranking for your target query — and the tool runs all the same metrics against each of them, then shows you a per-metric gap table.
The gap table is the thing. Your page at grade 14 against a competitor average of grade 9 is not a voice preference, it's a ranking liability. A passive-voice rate of 28% against a competitor average of 9% explains why the copy feels heavier than theirs.
For each metric, the tool shows a red or green delta: "ahead" or "behind" relative to the competitor set. Fix the behinds in order.
The AI Rewrite Prompt
The last tab produces a prompt already populated with your current metrics, the competitor averages, and explicit target thresholds. Paste it into Claude or ChatGPT and ask for a rewrite that hits the targets without losing substantive claims.
The prompt respects the skill-level setting shared across all the jwatte.com tools. Advanced skips the explanations and gives you rewritten blocks ready to ship. Beginner includes the reasoning so you can build intuition. Intermediate is the default and produces standard length.
When to Run It
Every time you rewrite a page for SEO, run this tool on the old version before the rewrite and on the new version after. If you improved the keyword coverage but your grade level went up and passive voice rose, you made the page harder to read even though it looks more "expert" — that's a common rewrite regression and a reason some rewrites underperform the original.
Run it on competitor pages you admire to reverse-engineer why they feel effortless.
Run it on your own archive quarterly. Pages written in different years rarely share a voice unless someone is actively editing toward one. The tool will tell you exactly where the drift lives.
Honest Limits
The passive-voice detector is a regex. It catches the obvious ("was sent," "has been made") and misses a few rare constructions. The Flesch formulas are linguistics from 1948 and they don't know that "laundromat" is three syllables even though the algorithm counts two. These are approximations, not diagnoses.
The tool reads static HTML. JavaScript-rendered content is invisible to it. If your page shell has 80 words of body copy and the rest is injected by React after first paint, the metrics will reflect the shell.
Use it as a yardstick, not an oracle. When it disagrees with an experienced editor, trust the editor.
How This Fits the Methodology
The voice analysis here mirrors the way the $20 Agency methodology thinks about content: measurable signals, published rubrics, nothing hand-wavy. Chapters 8 (Blogs 24/7 Sales Team) and 9 (Pillar-Cluster Model) walk through the editorial rules this tool quantifies. If you're operating a network of properties, the voice-consistency pass becomes a network-level audit — see $100 Network Chapter 16 on LLM-optimized content for how confident, direct prose gets cited more often by AI answer engines. If this is the first page on a brand-new site, $97 Launch covers the foundational voice work you do before you even start measuring.