HomeAI Detector › Score Your Blog Post for AI

Score your blog post for AI detection — a pre-publish E-E-A-T check.

Paste the draft. Get a 0-100 Authenticity Score in roughly thirty seconds, plus a sentence-level colour map that shows which paragraphs read AI to Google's helpful-content classifier and which read in your own voice. Built for bloggers and content writers checking listicles, how-to guides, reviews, opinion pieces, and evergreen long-form before they ship to WordPress, Ghost, Webflow, or Substack. AdSense-aware framing, keywords preserved across edits, no signup for the first scan.

Score my blog post free See pricing
0-100 score in 30 seconds Sentence-level highlights AdSense-aware WordPress + Ghost ready
The 2026 reality

Why publishing without a pre-flight is now expensive.

In 2023 you could publish an unedited ChatGPT draft and rank for a long-tail query inside two weeks. That stopped working in 2024 and is actively penalised in 2026. For bloggers there are three overlapping risks; pre-publish scoring addresses the prose-pattern layer of all three before the post leaves your editor.

The March 2024 core update and the 2025 helpful-content refinements

Google folded the helpful-content classifier into core ranking in March 2024 and added explicit scaled-content-abuse signals. The 2025 refinements added pattern detection for AI-typical prose. Sites that leaned on unedited drafts lost a meaningful share of organic traffic across the next two core updates, with reports clustering in the 30 to 60 percent range. Scoring before you publish is the cheapest way to catch the prose patterns that cause the drop.

AdSense and Publisher Network reviews

The monetization risk is separate from the ranking risk. AdSense is increasingly catching low-value scaled content and disabling ad serving on a per-page or per-site basis. The risk concentrates around thin posts, low-effort niche sites, and YMYL topics. A pre-publish score does not make weak content valuable, but it does prevent the prose pattern from being the thing that triggers a review on otherwise-credible work.

Reader perception your subscribers can feel

For bloggers on Substack, Ghost, or a personal newsletter, the audience that pays attention is the list of people who already chose to hear from you. They opened your last thirty emails because the voice was yours. When the voice shifts toward something blander and more generic, they sense it before they can name it. The unsubscribe is a slow drift, not a single moment, and the score is your early warning that the drift has started.

The pre-publish workflow

Four steps from draft to publish-safe.

This is the routine content writers use before pushing a post to the CMS. For a 1,500-word piece, plan on 15 to 25 minutes start to finish. The point of sentence-level highlights is that you do not have to rewrite the whole post; you only have to rewrite the parts that read AI.

Step 1: Paste the post and read the score

Drop the full draft into the scoring panel. In about thirty seconds you get a 0-100 Authenticity Score plus a paragraph-by-paragraph breakdown. Save the baseline if the score is below your target so you can measure the delta after rewriting. For posts longer than 800 words on the free tier, paste the intro first, then the body, then the close; the score per section is more diagnostic anyway.

Step 2: Review the sentence-level highlights

Green sentences passed every signal. Yellow sentences tripped one or two and are usually fine on their own. Red sentences tripped three or more and are the highest-leverage edits. The colour map matters more than the headline number; a score of 65 with three red sentences is easier to fix than a score of 78 spread evenly across yellow.

Step 3: Revise the flagged sections in your own voice

Work through the red sentences one at a time. Ask whether you would actually say each one out loud to a reader. If not, rewrite it. Add a parenthetical aside, a specific number, a sentence fragment, an opinion you would only express to people who read you regularly. These are the cues AI prose strips. If you use the AI rewriter, stay on Balanced for SEO posts where keywords matter and Light for personal voice posts where cadence matters.

Step 4: Re-scan, verify keywords, ship to the CMS

Paste the revised draft back in. Target 75 for general blog content, 80 for YMYL niches and personal newsletters. Ctrl-F the target keyword phrase to confirm it is still in position. Then copy the cleaned text into WordPress, Ghost, Webflow, or Substack as usual. The free Chrome extension also lets you re-scan inline against the CMS block editor for quick checks during the final edit.

Score bands

What each range means for a blog post.

The bands below are calibrated against published blog content across the six formats we tuned the classifier on. Numbers shift slightly by format (review posts run a few points lower than how-to posts at the same prose quality) but the bands below are the working defaults to ship by.

80-100: Publish to your list

Reads as your voice. Helpful-content classifier sees no meaningful AI patterns. AdSense review risk on prose alone is low. Loyal subscribers will read it the way they read your last post. This is where any piece you bill as your own thinking should sit before it ships, and the floor for YMYL niches like personal finance, health, and law where reviewers spend the most time.

70-79: Acceptable for roundups and news reactions

Fine for a weekly links roundup, a guest post, or a quick news reaction where voice carries less weight than coverage. For a paid newsletter or a personal essay post, push to 80 by editing the red sentences. AdSense risk at this band is low but non-zero and tends to depend on the rest of the site's pattern, not the individual post.

55-69: Audience-trust risk zone

Loyal subscribers begin sensing AI prose at this band. A single post might not move the unsubscribe rate, but a run of three or four in this range over a month tends to. Helpful-content classifier risk rises sharply here, particularly on listicles and product reviews. Edit before publishing; this is the band where the rewriting pays the most per minute spent.

35-54: Reads heavily AI

High probability the post pulls down helpful-content signals on the entire site, not just the page. Subscribers who know your work will notice immediately. AdSense per-page review odds are meaningful. Do not publish at this score for any post that goes out under your byline. Restructure or rewrite, do not paraphrase your way out.

0-34: Likely unedited AI output

Almost certainly an unmodified ChatGPT or Claude draft. Platform AI tags are near-certain on Substack and Medium. AdSense risk is high on YMYL niches. The fix is to throw out the draft and start from a real observation, not to run it through a paraphraser. Paraphrased prose has its own detection fingerprint that current classifiers flag more aggressively.

Tuned by format

Calibrated for the formats bloggers actually publish.

ChatGPT defaults to slightly different patterns in each blog format. The scorer adjusts so the bands above match the register the format actually needs, instead of penalising natural genre conventions.

Listicle posts

"7 Ways to," "10 Best," "5 Things." ChatGPT writes the skeleton perfectly, which is the problem. The scorer down-weights the parallel sentence opener on every list item and flags it; one or two items written longer than the others, with the parallel opener broken, usually pulls the post out of the risk band.

How-to and tutorial posts

Numbered steps, prerequisites, screenshots. ChatGPT defaults to a uniform step style with the same imperative opener on every step ("First, open," "Next, click," "Then, save"). The scorer flags monotony in step openers and looks for first-hand friction notes a real practitioner would name. The helpful-content system weights demonstrated expertise on tutorials particularly heavily.

Product reviews and round-ups

The highest-risk category for AdSense review. ChatGPT defaults to an evenly-positive register that reads sales-y rather than reviewed. The scorer looks for at least one honest negative, a comparison number, or a real-world test result. Reviews without a single concrete dislike read AI to both classifiers and to attentive readers.

Opinion and thought-leadership pieces

Personal stance, argument, takeaway. ChatGPT hedges every claim and softens every position into a generic centrist register. The scorer rewards sharpened claim sentences and personal stake (a specific example, a number, a contrarian observation). Opinion pieces that read like everyone-could-have-written-this read AI.

Evergreen long-form and pillar pages

The 2,000 to 4,000 word pieces that anchor most content sites. ChatGPT structures these with template H2s and bullet-heavy bodies; the cumulative effect across sections is what trips the classifier. Score section by section instead of as one chunk so the band tells you which part of the post needs work, not just that something does.

Plans & pricing

Same scoring engine at every tier.

Free covers most solo bloggers shipping one post a week. Starter is the right step for two-to-four posts a week. Pro adds unlimited scans, longer per-scan caps, file upload, and the Chrome extension. Full details on the pricing page.

Free
$0/forever

 

Try blog-post scoring. No card, no email.
  • 3 scans per day
  • 5,000 chars per scan
  • Sentence-level highlights
  • Plagiarism risk bundled
Start free
Starter
$7.49/month

Billed $89.88/year — Save $30

For bloggers writing 2 to 4 posts a week.
  • 20 scans per day
  • 20,000 AI rewriter words/mo
  • Chrome extension
  • Email support
Get Starter
Business
$29.99/month

Billed $359.88/year — Save $120

For content teams and agencies.
  • 150,000 AI rewriter words/mo
  • REST API + WordPress hook
  • 5 team seats
  • White-label PDFs
Get Business

Yearly billing saves 25%. View full pricing →

E-E-A-T compliance

Experience, expertise, authority, trust — and the prose cues that signal them.

E-E-A-T is the lens Google's reviewers use to judge whether a page deserves to rank in 2026. The scorer cannot fabricate real experience for you; what it can do is keep the prose patterns from being the thing that flags an otherwise-credible post.

The first E is the hardest to fake

"Experience" was added in late 2022 specifically because expertise alone was being gamed by AI-assisted content. It means demonstrated first-hand encounter with the topic: a tool you actually used, a place you actually visited, a result you actually measured. ChatGPT cannot generate this; it can only mimic the surface. The scorer flags sections where the prose sounds experienced but the content is generic, so you can drop in the real anchor before publishing.

Expertise without the AI tells

Expertise is signalled by terminology, specificity, and willingness to take a position. ChatGPT's default register is the opposite: hedged, generic, evenly-balanced. The scorer flags claim sentences that read like a research summary and rewards posts that commit to a specific position rather than surveying three options.

Trust as a writing decision

Trust signals are the small honesty cues: a caveat that contradicts the headline, a disclosure of limitations, a "we tried this and it did not work" paragraph. ChatGPT removes these by default because they hurt the surface-level confidence of the prose. The scorer rewards their presence and flags posts that read suspiciously frictionless.

FAQ

Blog-post scoring frequently asked.

Does Google penalise AI-written blog posts under the helpful-content system?
Google's public line is that quality is what matters, not how a post was written. In practice the March 2024 core update folded helpful-content into core ranking and added explicit scaled-content-abuse signals; the 2025 refinements added AI-prose pattern detection. Posts that read formulaic or template-heavy lose rankings even when the facts are correct. Scoring the post before you publish surfaces the patterns so you can rewrite them, which is the part Google's classifier actually weighs.
What blog post score should I target before clicking publish?
For a paid newsletter or a personal blog where your voice is the product, target 80 or higher. For a listicle, roundup, or news reaction where voice matters less, 75 is acceptable. For YMYL niches (money, health, law) where AdSense reviewers spend the most time, push to 80 before publishing. Below 65 is the band where loyal subscribers start sensing AI prose and where AdSense per-page reviews cluster.
Is there a real AdSense risk on AI-feel blog content?
Yes. AdSense and other monetization programs are actively reviewing sites for low-value scaled content in 2026. Pages flagged as predominantly AI-generated with minimal editorial value can lose ad serving on a per-page or per-site basis. The risk is highest for thin posts, low-effort niche sites, and posts in YMYL categories. Pre-publish scoring catches the prose-pattern layer of that risk so the editorial value can stand on its own.
Which blog formats does the scorer work on?
Listicles, how-to and tutorial posts, product reviews and round-ups, opinion and thought-leadership pieces, and evergreen long-form. The classifier was tuned across the structural patterns each format uses. Length range that scores most reliably is 800 to 2,500 words. Longer posts should be split into the intro, the body, and the close so the score is calculated against a representative chunk.
Can I score a 2,000-word post on the free tier?
The free tier caps each scan at 5,000 characters, which covers about 800 words. For a 2,000-word post, split it into the intro, the body, and the closing and scan each separately. This is also useful for spotting which section carries the AI risk. Pro at $19.99 monthly ($14.99 effective on yearly) gives 10,000 characters per scan, enough for most full newsletters in one pass.
How does this fit into a WordPress, Ghost, or Webflow workflow?
Paste the draft into TextSight before you import it into the CMS. Run the scan, review the sentence highlights, rewrite the red ones, then re-scan to verify. Once the score clears your target, copy the cleaned text into WordPress, Ghost, Webflow, or Substack as usual. The free Chrome extension also works inline against the WordPress block editor and Ghost editor for quick re-scans during edits.
Do my old blog posts need re-scoring too?
If you wrote them with AI assistance and they live on a platform that retroactively scores content (Medium does this, Substack does not currently), a back-catalogue rescan is cheap insurance. For a blog on your own domain, platform tagging is not a risk, but AdSense per-page review and reader perception both are. Posts in the 35-54 band on TextSight read obviously AI to attentive readers.
Related

More for the blogger workflow.

Score your blog post before you publish. Ship in your own voice.

Free to try, no card. 0-100 score in thirty seconds, sentence-level highlights, AdSense-aware, keywords preserved.

Score my blog post free See pricing
E-E-A-T-aware · Helpful-content compliant · Pre-publish, not post-publish