Publish: Anthropic Claude Data Retention Policy After September 2025#4817
Conversation
✅ Deploy Preview for char-cli-web canceled.
|
✅ Deploy Preview for hyprnote ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
Grammar Check ResultsReviewed 1 article. Anthropic Claude Data Retention Policy 2026📄 The article is well-researched and clearly written with strong structure and compelling arguments about Anthropic's privacy policy changes. There are only minor punctuation issues, primarily around comma placement and British-style punctuation conventions. The content is accurate, timely, and effectively serves its purpose of informing readers about privacy considerations when choosing an AI provider. Found 5 issues: 💡 ClarityLine 11
This fragment lacks context as a standalone sentence. Consider connecting it to the previous sentence or expanding for clarity. 📋 Suggested fix (click to expand)🔹 Punctuation PlacementLine 19
Unnecessary comma after 'accounts' breaks the flow of the relative clause. 📋 Suggested fix (click to expand)Line 31
Extra space after comma before the opening bracket. Use British punctuation style with comma outside quotation mark. 📋 Suggested fix (click to expand)Line 87
The abbreviation 'i.e.' should be followed by a comma, and the phrase should have proper punctuation around the abbreviation. 📋 Suggested fix (click to expand)📋 OtherLine 73
This line appears to have no issues, but verify that 'Char' is consistently capitalized throughout (it is). 📋 Suggested fix (click to expand)Powered by Claude Haiku 4.5 AI Slop Check ResultsReviewed 1 article for AI writing patterns. Anthropic Claude Data Retention Policy 2026
Score: 25/50 (NEEDS REVISION)
This post opens strong with factual reporting on Anthropic's policy change, but deteriorates into marketing copy by the final third. The dominant AI patterns are: (1) Binary antithesis framing ('This isn't X, it's Y') used repeatedly to set up contrasts; (2) Metronomic rhythm, especially in lists and policy explanations where sentences land with predictable cadence; (3) Staccato fragments in the opening and throughout for artificial drama; (4) Clickbait/listicle heading formulas (interrogative 'How,' 'What About,' 'Worth Knowing'); (5) Conversational announcements that tell rather than show ('This is the context you need,' 'A few exceptions matter'); (6) Heavy product marketing disguised as analysis in lines 78-84, where the Char comparison becomes a sales pitch with emotional framing ('actual control,' 'actually trusts'). The middle sections (storage policy, API differences) are technical and clear, but the intro and outro read as LLM-generated marketing. The author should strip the binary antithesis setups, compress staccato lists into flowing prose, replace interrogative headings with direct labels, and remove the promotional framing in the final section. Total score: 25/50, indicating significant revision needed. The text needs to decide whether it's technical analysis or product promotion and commit fully to one voice. Found 35 issues (3 high, 7 medium, 25 low) HIGH — Obvious AI TellLine 31 —
Multiple AI patterns: (1) 'Anthropic's previous stance was clean' + 'That was the explicit promise' + 'In August 2025, that changed' = binary antithesis structure (old vs new). (2) Three-sentence lead-up before the fact, metronomic rhythm. (3) 'clean' is anthropomorphization (policies don't have moral cleanliness). Remove the setup; state the policy reversal directly. Suggested rewriteLine 91 —
(1) 'That's what actual control looks like' is marketing/testimonial framing that positions Char as the 'real' solution and Anthropic as fake. (2) 'It's not just... it's...' = antithesis. (3) 'defaults to whatever Anthropic decides next quarter' is sarcasm that undermines trust. The whole sentence is a value pitch for Char, not analysis. Suggested rewriteLine 93 —
This is a call-to-action disguised as a closing. 'your security team actually trusts' implies Anthropic is untrustworthy. The entire final section (from line 78 onward) reads like a product pitch, not technical writing. De-emphasize the Char promotion or remove entirely. Suggested rewriteMEDIUM — Likely AI PatternLine 11 —
Staccato fragments used for dramatic effect. 'For a while, that reputation was earned' is a setup for the pivot—classic LLM structure that signals 'things changed (read on).' Suggested rewriteLine 13 —
Metronomic rhythm: three short punchy sentences in succession (announcement, deadline, consequence). The final one-liner ('If you missed it, your data was in') reads like an ad tagline designed to create urgency. Suggested rewriteLine 33 —
Metronomic rhythm: 'If X... If Y...' parallel structure repeated for rhetorical emphasis. Also: 'nothing changed' and 'There's still' are conversational filler that could be compressed. Suggested rewriteLine 35 —
(1) 'The reaction was immediate' + 'flagged it as' = staccato drama setup before the payload. (2) Scare quotes around 'privacy pivot' to distance from the term while using it. (3) 'The opt-in training setting extends...' is wordy; compress. (4) 'can sit in Anthropic's training pipeline' is anthropomorphization (data doesn't 'sit'). Suggested rewriteLine 57 —
'explicitly excluded from the September 2025 consumer policy changes' is passive and wordy. Em-dash before 'that's covered' is a reframe device. Simplify by stating directly what applies to commercial customers. Suggested rewriteLine 71 —
(1) 'This isn't... But it's...' = binary antithesis setup. (2) 'and it matters when evaluating whether a company's stated privacy values match their actual behavior' is a listicle-style moral conclusion. Show the discrepancy; let the reader draw conclusions. Suggested rewriteLine 89 —
(1) 'policy trajectory gives you pause' is marketing/emotional framing. (2) 'you're not locked in' is conversational announcement. (3) Metronomic rhythm: four short sentences, each making a point. Compress and vary. Suggested rewriteLOW — Subtle but SuspiciousLine 15 —
Conversational announcement / filler. 'This is the context you need' is throat-clearing that tells the reader what they're about to get instead of showing them. It also patronizes by framing what follows as mandatory. Suggested rewriteLine 17 —
Heading uses a question format that reads as listicle/clickbait. Rephrase as a direct descriptor instead. Suggested rewriteLine 21 —
Announcement before showing content. 'A few exceptions matter' tells the reader what's coming instead of presenting it. Delete the preview. Suggested rewriteLine 23 —
Staccato fragment opening ('Policy violations.') followed by complete sentences. Inconsistent structure. Also: 'Trust and safety classification scores' is jargon-heavy filler. Suggested rewriteLine 25 —
Another staccato fragment opening. Matches the pattern established in line 14. Varies sentence construction unnecessarily. Suggested rewriteLine 27 —
Staccato fragment opening. The full sentence that follows makes the fragment redundant. Suggested rewriteLine 29 —
Heading uses passive voice and nominalized structure. More direct: identify what the change was (the toggle) rather than just labeling it as 'a change.' Suggested rewriteLine 37 —
Heading uses interrogative (how-to) listicle formula. More direct: name the action or state the outcome. Suggested rewriteLine 39 —
'To see where your account stands' is conversational announcement. Delete and lead with the path. Suggested rewriteLine 41 —
Metronomic rhythm: 'If the toggle is on... Turning it off...' parallel conditional structure. Compress and reduce redundancy. Suggested rewriteLine 43 —
'does not retroactively remove' is nominalized/passive. Also: 'once it's been incorporated' is anthropomorphization—data is stored, not 'incorporated' as if the model digested it. Simplify and remove the filler comparison setup. Suggested rewriteLine 45 —
Heading uses interrogative listicle formula with a qualifier ('for privacy-conscious users') that reads as marketing framing. Be direct about what the section covers. Suggested rewriteLine 47 —
Announcement before showing the specifics. 'meaningfully different' and 'notably stronger' are intensifiers that don't add information. Show the difference; don't announce it. Suggested rewriteLine 49 —
Metronomic rhythm: three sentences, each stating a fact about the API. Compress into tighter prose. 'automatically deleted' and 'never used' don't need separate sentences. Suggested rewriteLine 51 —
'But the default is 7 days' introduces a mild antithesis (what you can do vs what happens by default). Also: 'stricter than most providers' is significance inflation without specifics. Suggested rewriteLine 53 —
'For enterprise API customers who qualify' is passive/wordy. 'One caveat' is throat-clearing before the exception. Compress and vary structure. Suggested rewriteLine 55 —
Heading uses abbreviated list format without context. More descriptive: label these as commercial variants to distinguish from consumer. Suggested rewriteLine 59 —
Minor: 'unless legally required otherwise' is filler; 'unless required by law' is more direct. Suggested rewriteLine 61 —
Interrogative listicle heading. Replace with statement of what the section covers. Suggested rewriteLine 63 —
Minor: Structure is sound but 'certain features including web search' is vague. Specify web search or consolidate. Suggested rewriteLine 65 —
Structure is passable but 'and can be exercised' is awkward. Also: 'don't automatically come with a DPA' uses negation where a positive statement would be clearer. Rephrase. Suggested rewriteLine 67 —
Heading uses marketing language ('Worth Knowing About'). Direct: state what the section covers. Suggested rewriteLine 69 —
Minor wordiness. 'with near-perfect accuracy' is intensifier; 'with high fidelity' or 'accurately' is sufficient. Otherwise factual and clear. Suggested rewriteLine 73 —
Heading uses interrogative comparison formula. More direct: label what's being compared. Suggested rewriteLine 87 —
Minor: 'i.e.' is jargon. Use a dash or parenthetical. Also: 'lets you bring your own API key' is anthropomorphization ('lets' suggests permission-granting). Say 'accepts your Anthropic API key' or 'uses your API credentials.' Suggested rewritePowered by Claude Haiku 4.5 with stop-slop rules |
Blog Post Review: Humanizer + Stop-SlopFile: Humanizer Check (24 AI writing patterns)Score: 36/50 (PASS)
Strengths
HIGH severityLine 91 — Pattern #9 (Negative Parallelism) + Pattern #1 (Significance Inflation)
"That's what X looks like" is a telltale AI conclusion structure. Combined with "It's not just..." negative parallelism. Suggested rewrite:
Line 87 — Pattern #1 (Significance Inflation)
Overstating importance with "the distinction that matters." Suggested rewrite:
MEDIUM severityLine 57 — Pattern #13 (Em Dash Overuse)
Em dash creates "punchy sales" cadence. Suggested rewrite:
Line 35 — Pattern #5 (Vague Attribution)
Vague attribution without specific sources. Suggested rewrite:
Line 47 — Pattern #4 (Promotional Language)
"Notably stronger" is promotional. Suggested rewrite:
LOW severityLine 19 — Pattern #8 (Copula Avoidance)
Awkward construction. Suggested rewrite:
Line 19 — Pattern #25 (Hyphenated Word Pairs)
Overly consistent hyphenation. Suggested rewrite:
Lines 78-84 — Pattern #14 (Boldface Overuse)
Mechanical boldface in table headers. Consider removing bold or using regular sentence case. Stop-Slop Check (phrases, structures, rhythm)Score: 37/50 (PASS)
Strengths
HIGH severityLine 91 —
"That's what X looks like" is performative emphasis. "It's not just..." is antithesis framing. Suggested fix:
Line 93 —
"actually trusts" is a sales pitch with adverb ("actually"). Replace with factual benefit. Suggested fix:
Line 71 —
"This isn't X. But it's Y" is textbook antithesis. "it matters when evaluating" is telling-not-showing. Suggested fix:
MEDIUM severityLine 15 —
Tells the reader what they're about to learn instead of letting the article speak. Suggested fix: Delete this line. Line 47 —
Two adverbs ("meaningfully", "notably") add no value. Suggested fix:
Line 57 —
Passive voice + adverb ("explicitly"). Suggested fix:
Line 31 —
"stance was clean" and "That was the explicit promise" are framing/significance inflation. Suggested fix:
Line 67 —
"Worth Knowing About" is meta-commentary. Suggested fix:
LOW severityLine 21 —
"A few exceptions matter" is hand-holding. Suggested fix:
Line 87 —
Em dash reframe + "This is the distinction that matters" is significance inflation. Suggested fix:
Line 49 —
"Never" is a lazy extreme. Suggested fix:
Lines 17, 45, 61 —
Question-format headings are a mild listicle/clickbait pattern. Suggested fix: Reframe as statements (e.g., "Default Data Storage", "API Privacy Differences", "HIPAA and GDPR Compliance"). SummaryBoth checks pass the 35/50 threshold. The article is above average for AI-generated content: it has real opinions, specific facts, and avoids the worst AI vocabulary. The main areas for revision are:
|
Blog Post Review: Humanizer + Stop-SlopFile: Humanizer Check (24 AI writing patterns)Score: 37/50 (PASS)
Strengths
HIGH severityLine 91 — Pattern #9 (Negative Parallelism) + Pattern #1 (Significance Inflation)
"That's what X looks like" is a telltale AI conclusion structure. Combined with "It's not just..." negative parallelism. Marketing language positioning the product as philosophically aligned with the reader. Suggested rewrite:
Line 93 — Pattern #4 (Promotional Language)
Sales pitch tone with emphatic adverb "actually." Undermines the neutral investigative tone of the rest of the article. Suggested rewrite:
Line 87 — Pattern #1 (Significance Inflation)
Overstating importance with "the distinction that matters." Suggested rewrite:
MEDIUM severityLine 57 — Pattern #13 (Em Dash Overuse)
Em dash creates "punchy sales" cadence. Suggested rewrite:
Line 35 — Pattern #5 (Vague Attribution)
Vague attribution without specific sources. Generic expert citation. Suggested rewrite:
Line 47 — Pattern #4 (Promotional Language) + Pattern #7 (AI Vocabulary)
"Meaningfully" and "notably" are AI vocabulary adverbs. "Notably stronger" is promotional. Suggested rewrite:
Line 11 — Pattern #10 (Rule of Three)
Classic three-part list for rhetorical rhythm. Suggested rewrite:
Line 37–38 — Pattern #17 (Symmetrical Sentence Structure)
Parallel "If...If" construction feels formulaic. Suggested rewrite:
LOW severityLine 19 — Pattern #8 (Copula Avoidance)
Awkward construction. Suggested rewrite:
Line 19 — Pattern #25 (Hyphenated Word Pairs)
Overly consistent hyphenation. Suggested rewrite:
Lines 56, 13 — Pattern #19 (Emphatic Adverbs)
Emphatic adverbs that can be cut without losing meaning. Suggested rewrite:
Line 15 — Pattern #14 (Signposting)
Tells reader what they need rather than delivering it. Suggested rewrite: Delete this line or replace with: "Here's what changed." Lines 78–84 — Pattern #14 (Boldface Overuse)
Mechanical boldface in table headers. Consider removing bold or using regular sentence case. Stop-Slop Check (phrases, structures, rhythm)Score: 38/50 (PASS)
Strengths
HIGH severityLine 91 —
"That's what X looks like" is performative emphasis. "It's not just..." is antithesis framing. Both sentences are significance inflation. Suggested fix:
Line 93 —
Direct product pitch with adverb "actually." Reads as ad copy, not technical guidance. Suggested fix:
Line 71 —
Classic "This isn't X. But it's Y" antithesis. The long follow-up is telling-not-showing significance. Suggested fix:
MEDIUM severityLine 47 —
Two adverbs ("meaningfully", "notably") add no value. Suggested fix:
Line 57 —
Passive voice + adverb ("explicitly"). Suggested fix:
Line 31 —
"Stance was clean" and "That was the explicit promise" are framing/significance inflation. Suggested fix:
Line 15 —
Tells the reader what they're about to learn instead of letting the article speak. Suggested fix: Delete this line. Line 67 —
"Worth Knowing About" is meta-commentary in the heading. Suggested fix:
LOW severityLine 35 —
Staccato emphasis fragment. Metronomic pattern: short statement → quoted term → long data statement. Suggested fix:
Line 21 —
"A few exceptions matter" is hand-holding. Suggested fix:
Line 87 —
Em dash reframe + "This is the distinction that matters" significance inflation. Suggested fix:
Line 49 —
"Never" is a lazy extreme per stop-slop rules. Suggested fix:
Lines 17, 45, 61 —
Question-format headings are a mild listicle/clickbait pattern. Suggested fix: Reframe as statements (e.g., "Default Data Storage", "API Privacy Differences", "HIPAA and GDPR Compliance"). SummaryBoth checks pass the 35/50 threshold. The article is above average: it has specific facts, avoids the worst AI vocabulary, and delivers real information. The main areas for revision are:
Reviewed with humanizer (24 AI writing patterns) and stop-slop (phrases, structures, rhythm) |
Blog Post Review: Humanizer + Stop-SlopFile: Humanizer Check (24 AI writing patterns)Score: 36/50 (PASS)
The post avoids most of the 24 AI writing patterns. No issues with: copula avoidance, elegant variation, false ranges, emojis, curly quotes, collaborative artifacts, knowledge-cutoff disclaimers, sycophantic tone, excessive hedging, or generic positive conclusions. The main weaknesses are promotional language in the closing section and one vague attribution. HIGH severityLine 91 -- Pattern #4 (Promotional Language) + #9 (Negative Parallelism)
"That's what actual control looks like" is promotional framing. "It's not just a privacy toggle..." is a classic "not just X" negative parallelism. The tonal shift from investigative to sales pitch is jarring. Suggested rewriteYour data stays on your device. Anthropic's quarterly policy changes don't apply. Line 93 -- Pattern #4 (Promotional Language)
Sales pitch / testimonial framing. "Actually trusts" is loaded language that doesn't belong in a technical guide. Suggested rewriteDownload Char for macOS and connect your preferred AI provider. MEDIUM severityLine 35 -- Pattern #5 (Vague Attributions)
No names, no links. Vague attribution weakens credibility. Suggested rewriteName specific researchers or link to specific critiques, e.g. "EPIC's Alan Butler called it a 'privacy pivot' in a September 2025 statement." Line 35 -- Pattern #1 (Significance Inflation)
Dramatic framing without supporting detail. Who reacted? When? Suggested rewritePrivacy advocates and security researchers criticized the change within days. Lines 87-89 -- Pattern #4 (Promotional Language)
The entire closing section (lines 87-93) shifts from investigative journalism to product marketing. The comparison table (lines 76-84) is fine and speaks for itself, but the prose around it reads like ad copy. Suggested rewriteConsider letting the comparison table stand on its own, or add a brief factual note: "Char routes data through the API (7-day retention, no training). Notes stay on your device as plain markdown." Drop the marketing framing. LOW severityLine 11 -- Pattern #10 (Rule of Three)
Staccato three-item fragment for manufactured rhythm. Works contextually but is a mild AI tell. Line 15 -- Pattern #22 (Filler Phrase)
Conversational preview. Consider removing -- the content below establishes its own importance. Stop-Slop Check (phrases, structures, rhythm)Score: 37/50 (PASS)
The post is above the 35/50 threshold and reads better than most AI-generated content. Most sentences have clear actors, specific details, and minimal jargon. Issues are refinement-level, not structural. Banned Phrases & AdverbsLine 47 -- Adverb: "meaningfully"
Two adverbs in one sentence ("meaningfully," "notably"). Suggested fixThe consumer product and the API have different data policies. The API is stronger. Line 57 -- Adverb: "explicitly" + Passive voice
Who excluded them? Anthropic did. Suggested fixAnthropic excluded commercial customers from the September 2025 consumer policy changes. Line 93 -- Adverb: "actually"
Empty emphasis. Remove "actually". Suggested fixuse the AI provider your security team trusts. Structural IssuesLine 31 -- Dramatic Fragmentation
Fragment for emphasis. The previous sentence already states the policy clearly. Suggested fixRemove this sentence. "Anthropic's previous stance was clean: consumer chats would not be used for training" is sufficient. Line 35 -- Narrator-from-a-distance
Observing from above instead of showing who reacted. Suggested fixSecurity researchers and privacy advocates called it a "privacy pivot." Line 71 -- Binary Contrast + Meta-commentary
"This isn't X. But it's Y" is textbook binary contrast. "It matters when evaluating" is telling instead of showing. Suggested fixThe lawsuit shows how Anthropic has approached training data acquisition, which informs whether their stated privacy values match their behavior. Line 91 -- Telling Instead of Showing
Announcing profundity. Let the facts speak. Suggested fixRemove entirely or integrate into the preceding sentence. Rhythm PatternsLine 57 -- Em-dash reveal
Em-dash for dramatic pause. Suggested fixYour data is not used for model training. Anthropic's commercial terms cover this, and it is not subject to opt-in or opt-out toggles. Line 47 -- False agency
The toggle doesn't "return" you -- you return yourself. Suggested fixTurn it off to go back to 30-day retention. Positive Elements
Summary
Both checks pass. The article is well-researched and mostly well-written. The primary area for improvement is the closing section (lines 87-93), which shifts from factual analysis to marketing copy. Both checks independently flagged this as the weakest part. Secondary issues: one vague attribution (line 35), a few adverbs, and minor dramatic fragments. |
Blog Post Review: Humanizer + Stop-SlopAnthropic Claude Data Retention Policy After September 2025
Humanizer Check (24 AI writing patterns)Score: 36/50 (PASS)
The post is well-researched and specific with strong data points (dates, retention periods, the 60x metric). Main weaknesses are inconsistent voice (alternates between punchy editorial and corporate reporting), curly quotes (ChatGPT signature), and vague attributions. Specificity is the strongest dimension. HIGH -- Immediate fixesLine 31 -- Pattern 18: Curly Quotation Marks
Curly quotes appear throughout the post. This is a known ChatGPT signature. Replace all curly quotes with straight quotes. Suggested fixReplace all curly Line 35 -- Pattern 5: Vague Attributions
No specific sources named. "Security researchers and privacy advocates" is a weasel-word construction. Suggested fixName a specific person or organization, or link to a specific article. E.g.: "Security researchers flagged it as a 'privacy pivot'." Line 91 -- Pattern 4: Promotional Language + Pattern 9: Negative Parallelism
Marketing copy posing as a conclusion. "Not just X" is a negative parallelism. "Actual control" is a promotional assertion. Suggested fixLine 93 -- Pattern 4: Promotional Language
CTA with emotional manipulation ("actually trusts" implies Anthropic is untrustworthy). The adverb "actually" is an emphasis crutch. Suggested fixMEDIUM -- Should fixLine 11 -- Pattern 10: Rule of Three + Pattern 1: Significance inflation
Three-item list for rhetorical punch. "For a while, that reputation was earned" is a dramatic pivot setup. Suggested fixLine 15 -- Pattern 1: Significance inflation
"This is the context you need" overstates importance. Throat-clearing filler. Suggested fixDelete this line. The preceding paragraphs establish the stakes without announcement. Line 31 -- Pattern 1: Significance inflation
"That was the explicit promise. In August 2025, that changed." is a metronomic setup/reveal beat. Suggested fixLine 35 -- Pattern 10: Rule of Three
Staccato opener for dramatic effect, followed by unattributed claims. Suggested fixLine 89 -- Inconsistent voice / mechanical rhythm
Four sentences of similar length and cadence. "Gives you pause" is corporate-colloquial. "Starting over" is vague. Suggested fixLOW -- Nice to fixLine 19 -- Pattern 22: Filler phrase
"that are" is a clunky construction. Suggested fixLine 21 -- Throat-clearing
Announces what's coming instead of just listing it. Suggested fixLine 25 -- Pattern 25: Hyphenated Word Pairs
Consistently hyphenated compound modifiers throughout. Humans are less consistent. Minor tell. Line 71 -- Hedging
"Not X, but Y" setup. Hedges before making the point. Suggested fixStop-Slop Check (phrases, structures, rhythm)Score: 35/50 (PASS -- at threshold)
The post is informative and respects reader intelligence. No throat-clearing openers, no business jargon, no dramatic fragmentation. Primary weakness is a passive voice epidemic: the post systematically avoids naming Anthropic as the actor. "Conversations are saved" instead of "Anthropic saves conversations." A post criticizing Anthropic's privacy policies would hit harder if it named Anthropic as the actor in every sentence. Passive Voice Issues (systematic)These are grouped because they share the same fix pattern: name the actor.
Adverbs to Cut
Binary ContrastLine 31
Setup/reveal pattern. State the change directly. Telling Instead of ShowingLine 91
Matches the banned pattern "This is what X actually looks like" from phrases.md. Delete or replace with a specific statement about what the user controls. Strengths
Combined Recommendations (priority order)
|
Blog Post Review: Humanizer + Stop-SlopFile: Humanizer Check (24 AI writing patterns)Score: 37/50 (PASS)
HIGH — Clear AI PatternsLine 91 — Pattern #4 (Promotional Language) + #9 (Negative Parallelism)
"That's what actual control looks like" is announcement + evaluation. "It's not just a privacy toggle" is a negative parallelism ("not just X"). Together these are designed to land rhetorically, not to convey information. The claim about "defaults to whatever Anthropic decides next quarter" is speculation. Suggested rewriteLine 93 — Pattern #4 (Promotional Language)
"your security team actually trusts" is testimonial framing + marketing pitch language. Technical writing ends with the link, not a trust appeal. Suggested rewriteMEDIUM — Likely AI PatternsLine 11 — Pattern #25 (Hyphenated Word Pairs) + #10 (Rule of Three)
Three-item fragment list used for dramatic effect. "privacy-conscious" and "safety-first" are consistently hyphenated (humans are sloppier). The fragment list is manufactured rhythm. Suggested rewriteLine 13 — Pattern #4 (Promotional Language)
Marketing-copy dramatization. A technical writer would state the policy plainly. Suggested rewriteLine 47 — Pattern #7 (AI Vocabulary)
"meaningfully" and "notably" are AI-favored intensifiers that add no specificity. Suggested rewriteLine 71 — Pattern #9 (Negative Parallelism)
Binary antithesis ("This isn't about X. But it is about Y."). The core claim can be stated in one sentence. Suggested rewriteLine 89 — Pattern #4 (Promotional Language)
Conversational marketing language with metronomic rhythm (three short declarative sentences about freedom). Tighten to functional benefits. Suggested rewriteLOW — Subtle PatternsLine 15 — Pattern #22 (Filler Phrase)
Conversational announcement that doesn't add information. The content that follows stands on its own. Suggested rewriteLine 19 — Pattern #22 (Filler Phrase)
"before being permanently deleted" is slightly redundant (deletion is inherently permanent). Suggested rewriteLine 21 — Pattern #22 (Filler Phrase)
Unnecessary setup phrase before introducing a list. Suggested rewriteLine 31 — Pattern #1 (Emphasis on Significance)
Metronomic three-sentence rhythm ("clean stance" → "explicit promise" → "that changed"). The emphasis is manufactured. Suggested rewriteLine 35 — Pattern #5 (Vague Attribution)
Vague attribution (no named researchers). Borderline acceptable for a blog post. Line 43 — Anthropomorphization
"Doesn't unlearn" is anthropomorphization. Models are trained/retrained, they don't "learn/unlearn." Suggested rewriteLine 87 — Pattern #25 (Hyphenated Word Pairs)
Consistent hyphenation throughout. Also "de-identified" (line 33), "privacy-conscious" (line 11, 45). Humans hyphenate inconsistently. Stop-Slop Check (phrases, structures, rhythm)Score: 35/50 (BORDERLINE — at threshold)
Banned PhrasesLine 15 — Throat-clearing opener
Announces what follows instead of stating it. Cut the sentence. Line 47 — Filler adverbs
Adverbs doing vague emphasis work. Delete both. Line 71 — Meta-commentary
Meta-commentary about why information matters, rather than letting the information speak. Cut and state the point. Line 91 — Telling instead of showing
"This is what X actually looks like" is a banned phrase pattern. State the benefit directly. Line 93 — Adverb
"Actually" adds nothing. Delete. Structural ClichesLine 71 — Binary contrast
"Not X. But Y." is a mechanical contrast. State Y directly. Line 91 — Binary contrast
"Not just X" is an additive hedge from the binary contrast patterns. Drop the negation. Line 31 — Dramatic fragmentation
Staccato emphasis. Two dramatic sentence fragments building a reveal. Line 13 — Dramatic fragmentation
Punchy one-liner for dramatic effect. Rhythm PatternsLine 11 — Three-item list
Three items. Use two or one. Line 89 — Metronomic endings
Three consecutive sentences of similar length with punchy endings. Break the pattern. Line 35 — Staccato fragmentation
Short punchy sentence followed by explanation. Passive voice hides the actor. Suggested fixStrengths
SummaryThe factual core of this post (lines 17–65) is solid — specific dates, numbers, and policy details presented directly. The opening (lines 11–15) and closing (lines 87–93) sections carry the most AI patterns: dramatic fragmentation, promotional language, binary contrasts, and manufactured emphasis. The middle technical sections need only minor cleanup (adverbs, filler phrases). Strongest recommended changes:
|
Blog Post Review: Humanizer + Stop-SlopReviewed Humanizer Check (24 AI writing patterns)Score: 40/50 (PASS)
The post is largely clean with strong voice, specific details, and good rhythm variation. Most of the 24 patterns are absent. Issues found are minor. HIGHLine 91 -- Pattern #9: Negative Parallelism
"It's not just X; it's Y" construction paired with testimonial framing ("That's what actual control looks like"). Suggested rewriteLine 93 -- Pattern #4: Promotional Language
Marketing CTA with emotional appeal ("actually trusts"). Suggested rewriteMEDIUMLine 15 -- Pattern #19: Collaborative Communication Artifact
Service-y framing ("this is the context you need"). Suggested rewriteLine 67 -- Pattern #1: Undue Emphasis on Significance
"Worth Knowing About" inflates significance rather than describing content. Suggested rewriteLine 25 -- Pattern #25: Hyphenated Word Pairs
"back-end" is inconsistently hyphenated; "backend" is standard in tech writing. Suggested rewriteLOWLine 19 -- Pattern #22: Filler / Awkward Construction
Grammatically awkward "that are" construction. Suggested rewritePatterns NOT Found (Good Signs)
Stop-Slop Check (phrases, structures, rhythm)Score: 37/50 (PASS)
Above the 35/50 threshold. Main issues are adverb creep, a few passive voice instances, and two "telling instead of showing" patterns. The factual grounding and specificity are strong. Banned PhrasesLine 47 -- Adverbs: "meaningfully," "notably"
Both adverbs add no information. Suggested fixLine 71 -- Adverb: "directly"
"directly" is filler. Suggested fixLine 91 -- Telling instead of showing: "actual"
"actual" is an emphasis crutch; "that's what X looks like" is a banned pattern. Suggested fixLine 93 -- Adverb: "actually"
"actually" is performative emphasis. Suggested fixStructural ClichesLine 21 -- Meta-commentary
"That's the standard case" is throat-clearing before the point. Suggested fixLine 35 -- Narrator-from-a-distance
Floating observation; the next sentence already names the actors. Suggested fixDelete this sentence. The next sentence ("Security researchers and privacy advocates flagged it...") already delivers the point. Line 71 -- Binary contrast
"This isn't X. But it's Y." binary antithesis. Suggested fixPassive VoiceLine 19 -- "is flagged"
Hides who does the flagging. Suggested fixLine 57 -- "were explicitly excluded"
Adverb + passive voice. Suggested fixLine 43 -- "does not retroactively remove"
Suggested fixRhythm IssuesLine 89 -- Metronomic short sentences
Four sentences of similar length building to a punchy conclusion. "policy trajectory gives you pause" is business jargon. Suggested fixSummary
Overall verdict: PASS with minor revisions recommended. The post is well above the revision threshold on both checks. The factual content is strong and specific. The main areas for tightening:
|
Blog Post Review: Humanizer + Stop-SlopFile: Humanizer Check (24 AI writing patterns)Score: 40/50 (PASS)
The post is strong on specificity (concrete dates, numbers, source links) and natural tone (contractions, direct reader address). Remaining issues are minor mechanical patterns. HIGHLine 91 — Pattern #9: Negative Parallelism
"It's not just X" is a classic AI negative parallelism. Combined with testimonial framing ("That's what actual control looks like") and a sarcastic jab. Suggested rewriteMEDIUMLines 78-84 — Pattern #14: Overuse of Boldface
Every label in both columns and rows is bolded, which reads as over-formatted. Suggested rewriteRemove boldface from row labels; keep only column header bold. Line 87 — Readability
"i.e." needs parentheses or commas for readability. Suggested rewriteLOWLine 45 — Pattern #25: Hyphenated Word Pair Overuse
Consistent technical hyphenation can read as AI-perfect. Minor tell. Lines 23-27 — Structural repetition
Identical fragment-then-sentence structure repeated three times. Vary the pattern. Stop-Slop Check (phrases, structures, rhythm)Score: 35/50 (BORDERLINE — NEEDS REVISION)
The first half is informative and tight. The second half (from line 67 onward) shifts into promotional voice with binary contrasts and metronomic rhythm. Banned PhrasesLine 13 — Adverb: "quietly"
Editorial interpretation disguised as fact. Let the timing speak for itself. Suggested fixLine 47 — Adverb stacking: "meaningfully different" + "notably stronger"
Two adverbs adding emphasis without information. Suggested fixLine 57 — Adverb: "explicitly excluded"
"Explicitly" is filler. Suggested fixLine 91 — Telling instead of showing
Announces significance rather than demonstrating it. "This is what X actually looks like" is a banned pattern. Suggested fixDelete or replace with: "You control your data. You choose your provider." Line 93 — Telling instead of showing: "actually trusts"
"Actually" is a banned adverb. The whole phrase is marketing copy. Suggested fixStructural ClichesLines 11-13 — Dramatic contrast setup
Setup → reversal structure with staccato fragments and "Then" signaling a scripted turn. Suggested fixCompress the setup. Start closer to the change itself. Line 71 — Binary contrast: "not X, but Y"
Negation-then-assertion pattern. Suggested fixLine 91 — Binary contrast: "not just X"
Classic "not X, it's Y" structure. Suggested fixState the positive claim directly without the negation. Rhythm PatternsLine 11 — Three-item list
Rule of three. Suggested fixUse two items: "Constitutional AI and no training on customer data." Lines 13 — Metronomic rhythm
Progressively shorter sentences manufacturing urgency. Suggested fixLines 89 — Metronomic rhythm (4 sentences building to a punchy closer)
Each sentence is a reassurance building to a slogan. Suggested fixPassive VoiceLines 23-27 — Multiple passive constructions
Name Anthropic as the actor: "Anthropic keeps inputs and outputs for 2 years." Meta-CommentaryLine 15 — Throat-clearing
Announces what the reader should think. Suggested fixDelete, or: "This context matters if you're evaluating AI providers." Line 21 — Throat-clearing
Narrates the article's own structure. Suggested fixDelete "That's the standard case." Start with: "Exceptions:" SummaryThe article scores well on the humanizer check (40/50 — PASS) due to strong specificity, concrete sourcing, and natural contractions. The stop-slop check (35/50 — BORDERLINE) flags more issues, primarily: adverb use ("quietly," "meaningfully," "notably," "explicitly," "actually"), binary contrast structures in the closing sections, metronomic rhythm patterns, passive voice hiding Anthropic as actor, and meta-commentary that narrates the article structure. The closing three paragraphs (lines 87-93) are the weakest, shifting from factual reporting to promotional framing. Top 5 changes for maximum impact:
|
ComputelessComputer
force-pushed
the
blog/anthropic-data-retention-policy-1774863673101
branch
from
17bd997 to
d318223
Compare
Blog Post Review: Humanizer + Stop-SlopFile: Humanizer Check (24 AI writing patterns)Score: 37/50 (PASS)
Strengths
HIGH severityLine 91 -- Pattern #9 (Negative Parallelism) + Pattern #1 (Significance Inflation)
"That's what X looks like" is a telltale AI conclusion structure. Combined with "It's not just..." negative parallelism. Suggested rewrite:
Line 93 -- Pattern #4 (Promotional Language)
Sales pitch tone with emphatic adverb "actually." Undermines the neutral investigative tone of the rest of the article. Suggested rewrite: Line 15 -- Pattern #27 (Persuasive Authority Tropes)
Authority framing -- tells the reader what they need rather than letting the content speak for itself. Suggested rewrite:
MEDIUM severityLines 17, 37, 45, 61, 67, 73 -- Pattern #17 (Title Case in Headings) All headings use Title Case, which is the single biggest AI tell in the article. Convert to sentence case.
Line 47 -- Pattern #7 (AI Vocabulary)
"Meaningfully" and "notably" are high-frequency AI adverbs. Suggested rewrite:
Line 35 -- Pattern #5 (Vague Attribution)
Vague attribution without specific sources. Suggested rewrite:
Line 11 -- Pattern #10 (Rule of Three)
Classic three-part list for rhetorical rhythm. Suggested rewrite:
Line 57 -- Pattern #14 (Em Dash Overuse)
Em dash creates "punchy sales" cadence. Suggested rewrite:
LOW severityLine 21 -- Pattern #23 (Filler Phrase)
"A few exceptions matter" is filler. Suggested rewrite:
Line 35 -- Pattern #1 (Significance Inflation)
Inflated significance framing. Merge with next sentence. Suggested rewrite:
Line 67 -- Pattern #27 (Persuasive Authority Tropes)
"Worth Knowing About" is authority framing in a heading. Suggested rewrite:
Lines 78-84 -- Pattern #15 (Boldface Overuse)
Consider using plain column headers or minimal bold. Stop-Slop Check (phrases, structures, rhythm)Score: 37/50 (PASS)
Strengths
HIGH severityLine 91 --
"That's what X looks like" is performative emphasis. "It's not just..." is antithesis framing. Suggested fix:
Line 93 --
"actually trusts" is a sales pitch with adverb. Replace with factual benefit. Suggested fix: Line 71 --
"This isn't X. But it's Y" is textbook antithesis. "it matters when evaluating" is telling-not-showing. Suggested fix:
MEDIUM severityLine 15 --
Tells the reader what they're about to learn instead of letting the article speak. Suggested fix: Delete this line. Line 47 --
Two adverbs ("meaningfully", "notably") add no value. Suggested fix:
Line 57 --
Passive voice + adverb ("explicitly"). Suggested fix:
Line 13 --
"Quietly" is an adverb that anthropomorphizes the announcement. Suggested fix:
Line 35 --
Vague declarative. Show the specific reaction instead. Suggested fix: Merge with the next sentence: "Security researchers and privacy advocates called it a 'privacy pivot.'" Line 57 --
Em-dash. Replace with period or comma. Suggested fix:
Line 67 --
"Worth Knowing About" is meta-commentary. Suggested fix:
LOW severityLine 21 --
"A few exceptions matter" is hand-holding. Suggested fix:
Line 19 --
Double passive. Suggested fix:
Line 23 --
Hidden actor. Suggested fix:
SummaryBoth checks pass the 35/50 threshold. The article is above average for AI-generated content: it has real opinions, specific facts, linked sources, and avoids the worst AI vocabulary. The main areas for revision are:
|
ComputelessComputer
deleted the
blog/anthropic-data-retention-policy-1774863673101
branch
ComputelessComputer
restored the
blog/anthropic-data-retention-policy-1774863673101
branch
Blog Post Review: Humanizer + Stop-SlopFile: Humanizer Check (24 AI writing patterns)Score: 31/50 (NEEDS REVISION)
Strengths
HIGH Severity
MEDIUM Severity
LOW Severity
Voice/Rhythm Notes
Stop-Slop Check (phrases, structures, rhythm)Score: 36/50 (PASS)
Issues by CategoryFiller Adverbs (7 instances)
Passive Voice (18 instances — biggest issue)
Telling Instead of Showing (3 instances)
Binary Contrast / Marketing Framing (3 instances)
Metronomic Rhythm (2 instances)
Combined SummaryThe article is well-researched with strong factual content, specific dates, and good sourcing. The middle sections (storage policy, API differences, enterprise tiers) are solid technical writing. Top 5 changes for maximum impact:
Combined Score: 67/100 — The factual core is strong. Fix passive voice, cut adverbs, and tone down the closing promotional framing. |
ComputelessComputer
deleted the
blog/anthropic-data-retention-policy-1774863673101
branch
2 participants
Article Ready for Publication
Title: Anthropic Claude Data Retention Policy After September 2025
Author: Harshika
Date: 2026-03-30
Category: Guides
Branch: blog/anthropic-data-retention-policy-1774863673101
File: apps/web/content/articles/anthropic-data-retention-policy.mdx
Auto-generated PR from admin panel.