Pre-scan ad creative, long-form sales pages, VSL scripts, landing page heroes, and cold email sequences before they ship into paid media. Sentence-level highlights surface the templated openers, generic urgency, and stock ROI claims that get flagged in client review. Built for freelance direct-response copywriters, in-house brand teams, and agency creative pods. Free to try. No card.
Freelance direct-response copywriters running their own roster, in-house brand copywriters working a single voice across a media plan, and agency creative pods shipping ad sets and sales pages across multiple accounts share the same need: a fast pre-launch scan that flags AI-shaped copy before a reviewer, a client, or a quality score does.
Copywriting is one of the most exposed disciplines to generative-AI review. Meta and Google reviewers tightened ad-creative checks through 2025. Brand leads run their own Originality.ai or Copyleaks pass on agency drafts before signing off on a launch. The realistic 2026 workflow uses AI assistance and pre-scans the output before it goes anywhere near paid spend.
Five to fifteen pieces a week across two to five clients. Pro at $14.99 a month yearly gives unlimited scans, 10,000 character pastes (enough to fit a long-form sales page in one go), and 90-day history covering roughly the length of a single client retainer. The integrated AI rewriter handles the stubborn paragraphs that flag every time without forcing a rewrite of a tested control.
One or two writers running brand voice across a media plan, with ad sets and landing pages going out weekly. Pre-scanning every piece before it reaches the brand-review meeting cuts the AI-pushback loop with brand leadership. The PDF export keeps a defensible record per launch, useful when leadership later asks how creative is handling AI quality control across the calendar.
Five to thirty copywriters shipping fifty or more deliverables a month across multiple client accounts. Business at $29.99 a month yearly unlocks five seats with shared scan history, REST API access for workflow automation, an audit log, and white-label PDFs branded to the agency. Direct-response agencies running a minimum Authenticity Score on every piece of paid creative usually settle on Business inside their first quarter.
A Facebook ad and a long-form sales page are not the same animal. Each copy genre has its own register, its own paraphrase density, and its own false-positive risk. Read the score in context of the format rather than chasing one number across every piece of creative.
Short-form, three to seven lines of body plus a headline. Chunk size is below the classifier's reliable band, so single ads score noisily. The realistic move is to scan the full ad set as one paste (five to ten variants together) so the model has enough signal to score consistently. Healthy scores on hand-written direct-response ads run 70 to 85. Generic urgency openers and unspecified ROI claims are the recurring flags.
Ninety-character headlines and ninety-character descriptions. Even shorter than Meta. Scan the entire campaign of fifteen headline variants together rather than scoring one headline at a time. The classifier looks at the joint distribution, which is the only stable read on copy this compressed.
Spoken-word script for a video sales letter, typically twelve to twenty-five minutes of read time. Scores well when the script preserves spoken cadence (contractions, sentence fragments, one-word lines) and scores poorly when it slips into written-prose rhythm. Healthy scores run 75 to 90 on a script written from a real interview rather than an outline.
Three to ten thousand words of structured sales argument with hook, story, offer, proof, and close. The headline number averages across the whole page, which masks specific weak slots. Scan the full page first, then re-scan the hook and the close separately, because those are the sections that drift into stock phrasing under deadline pressure.
Short-form pieces below the reliable chunk band. Scan the full sequence (five to seven emails) as one paste rather than scoring one email at a time. Cold-email opening lines are the highest-risk slot because they default to the templated "I noticed that you" register that LLMs produce by default.
Headline, sub-headline, three benefit lines, and a CTA. Structurally similar across the entire SaaS category by convention, so scoring is volatile on the chunk alone. Treat the headline number as advisory and read the sentence highlights. Specific concrete vocabulary lifts these pieces faster than rewriting the whole hero block.
Pro at $19.99 a month standard, $14.99 a month on yearly, is the right fit for freelance and in-house copywriters. Business at $39.99 a month standard, $29.99 a month on yearly, fits agency creative pods scanning fifty or more deliverables a month. Full details on the pricing page.
Billed $89.88/year — Save $30
Billed $179.88/year — Save $60
Billed $359.88/year — Save $120
Yearly billing saves 25%. View full pricing →
Copywriting frameworks are not the enemy. A framework filled in with generic language is. AIDA, PAS, and 4Ps produce templated structural moves that overlap heavily with what an LLM defaults to when prompted for a sales page. The fix is to keep the framework intact and load the body with specifics.
Attention, Interest, Desire, Action still works, but the Attention slot is the one that flags hardest. Generic openers (Imagine, Picture this, Unlock, Discover) read as templated to both reviewers and classifiers. Replace the slot with a specific number, a named customer, or a verbatim quote from research, and the framework slot still does its job without the AI fingerprint.
Problem, Agitate, Solve scores best when the Agitate slot uses concrete customer-voice phrasing instead of generic "you are tired of struggling with" language. The structural move is fine. The vocabulary is the giveaway. Sentence highlights show which Agitate lines dropped into the templated band so you can swap in research-derived phrasing surgically.
Promise, Picture, Proof, Push relies on parallel structure across the four slots. LLMs produce parallel structures by default, so a Four Ps page often flags hard if the slots are all the same sentence length and cadence. Vary the rhythm between slots (short, long, fragment, long), add specific proof figures, and the framework reads human while still delivering the rhythm readers expect.
The recurring offending openers are Unlock, Discover, Transform, Elevate, Revolutionise, and Master. They flag in headline scans because they sit in the templated band of LLM defaults. Healthier openers lead with a specific noun, a number, or a verbatim quote pulled from research. The discipline is to write the opener last, after the offer and proof are settled, so it can carry specific content rather than acting as a generic warm-up.
Brand voice and detector pass-through line up more often than they fight. Distinctive cadence, varied sentence length, and concrete vocabulary all read on-brand to a brand lead and read human to the classifier. The pieces that flag are usually the ones where a writer dropped into the neutral SaaS register to push through a deadline.
Copy built from research interviews and customer-support transcripts almost always scores well because the vocabulary is specific to the audience rather than generic to the category. The discipline of pulling verbatim phrases into the draft does double duty: it lifts conversion and lifts the Authenticity Score at the same time.
A brand style guide that requires product-specific terminology over generic SaaS language is also requiring prose that reads less templated. The two requirements collapse into one. Following the brand vocabulary lifts the detector score as a side effect of doing the brand work properly.
Rare, but it happens. Some brand voices require a formulaic sign-off or boilerplate disclosure that flags every time. Flag this in the kickoff so the client knows the boilerplate paragraph will run lower than the body, then defend it on its merits rather than rewriting required copy.
The score is the diagnostic, not the goal. Rewriting purely to lift the number tends to flatten the voice and kill conversion-tested phrasing. Use the sentence highlights to find specific lines that drift into stock phrasing, rewrite those, and let the headline number land where it lands. The voice survives and the deliverable still passes.
A pre-handoff scan is the move that keeps brand leads from running their own check and asking awkward questions on the review call. Build the scan into the delivery ritual so every piece of copy ships with a screenshot of the score and the date in the email or Slack thread.
Most direct-response agencies settle on an Authenticity Score floor of 75 or 80 for paid creative. Communicate the policy in the kickoff so a writer sending a 72 knows to revise before submitting. The number itself matters less than the consistency. A stable floor across every deliverable signals to the brand lead that quality control runs the same way every time.
Paste a screenshot of the score and the timestamp into the body of the delivery message, or attach the PDF export on Pro and Business. The point is to pre-empt the conversation. Once a brand lead receives one delivery with the scan attached, they stop running their own check on every piece because the trust is established.
If the brand requests substantive edits, rescan the revised piece and reattach the report. This catches the case where a revision accidentally drops the score by introducing a templated passage to fix a different issue. Keeping receipts on the revision cycle prevents a late-stage flag from coming out of nowhere on a piece that scanned clean originally.
The audit log on the Business tier shows which copywriter scanned which deliverable, with timestamps, and which scans were exported as PDFs. Useful for quarterly client reviews where the agency wants to demonstrate consistent AI-quality control across the creative pod rather than relying on individual judgment.
A high score is not the win condition on its own. The win condition is copy that converts and does not get flagged. Use the score as one diagnostic alongside open rate, click rate, and conversion rate rather than chasing the headline number at the expense of a tested control.
On a control that is already converting, the right AI rewriter setting is Light. It preserves sentence structure and only adjusts cadence, which means the tested hook, offer, and close stay intact while the AI fingerprint comes off. Maximum mode is reserved for stock phrasing that was templated to begin with and was never a tested element.
When generating A/B variants from a winning control, scan each variant before launch and discard anything that scores below the floor. This prevents a generative-AI variant from polluting the test cell and giving you a false read on whether the new hook is the issue or the AI residue is.
Subject lines and preheaders are the highest-AI-risk slots in an email sequence because they default to templated curiosity ("You will not believe what happened next") and templated urgency ("Last chance, ending tonight"). Scan subject-line variants in batches of ten or twenty so the chunk is big enough to score reliably, then pick the highest-converting one that clears the score floor.
Landing page heroes correlate strongly with hero-level conversion rate. A hero that reads generic ("Transform your business with our platform") sits low on both score and conversion. A hero with a specific number and a named customer ("Cut onboarding from twelve days to three for Series B SaaS") sits high on both. The detector is reading the same flatness signal the visitor reads.
A single percentage is not a fix path. The TextSight result panel shows which sentences reacted and why, with paragraph-level rollups for longer pieces, so you can edit specific lines instead of rewriting the whole page.
Every sentence is colour-coded by its own AI-likeness score. Red sentences clustered in one paragraph are a stronger signal than scattered yellows. Scattered yellows in otherwise structured prose often just mean a stock transitional phrase. You read the pattern, not just the headline number.
Longer pieces get paragraph-level rollups so you can see which paragraph is dragging the headline score. On sales pages the hook and the close are the usual suspects. On VSL scripts the bridge between problem and solution is the recurring weak slot. Targeting the lowest paragraph first is the fastest way to lift the page.
Perplexity is how predictable your word choices are to a language model. Low perplexity reads AI-like. The score is shown per-sentence on Pro, which is the diagnostic context you need to decide whether a flag is real AI residue or a particularly well-rehearsed product description.
Burstiness is how much sentence length and structure vary across the piece. ChatGPT defaults to uniform medium-length sentences. Real human copywriting has bursty rhythm: one short line, one long, one fragment. Low burstiness across an entire sales page is the classic AI fingerprint and the one experienced brand leads learn to spot first.
More for copywriters.
The full content-writer workflow with delivery-attached scans and brand voice defence.
For writers →How agency teams running fifty-plus pieces a month build the scan into their QA workflow.
For agencies →Light, Balanced, and Maximum modes for fixing flagged passages without losing voice.
Read the guide →Free, Starter, Pro, Business. Yearly billing saves 25%. Solo to agency tiers.
See pricing →Free to try. No card. Pro at $14.99 a month on yearly for solo copywriters; Business at $29.99 a month on yearly for agency creative pods.