| 100.00% | Adverbs in dialogue tags | Target: ≤10% dialogue tags with adverbs | | totalTags | 4 | | adverbTagCount | 0 | | adverbTags | (empty) | | dialogueSentences | 13 | | tagDensity | 0.308 | | leniency | 0.615 | | rawRatio | 0 | | effectiveRatio | 0 | |
| 95.14% | AI-ism adverb frequency | Target: <2% AI-ism adverbs (58 tracked) | | wordCount | 1029 | | totalAiIsmAdverbs | 1 | | found | | | highlights | | |
| 100.00% | AI-ism character names | Target: 0 AI-default names (17 tracked, −20% each) | | codexExemptions | (empty) | | found | (empty) | |
| 100.00% | AI-ism location names | Target: 0 AI-default location names (33 tracked, −20% each) | | codexExemptions | (empty) | | found | (empty) | |
| 65.99% | AI-ism word frequency | Target: <2% AI-ism words (290 tracked) | | wordCount | 1029 | | totalAiIsms | 7 | | found | | | highlights | | 0 | "rhythmic" | | 1 | "reminder" | | 2 | "racing" | | 3 | "raced" | | 4 | "quickened" | | 5 | "scanning" | | 6 | "flickered" |
| |
| 100.00% | Cliché density | Target: ≤1 cliche(s) per 800-word window | | totalCliches | 0 | | maxInWindow | 0 | | found | (empty) | | highlights | (empty) | |
| 100.00% | Emotion telling (show vs. tell) | Target: ≤3% sentences with emotion telling | | emotionTells | 1 | | narrationSentences | 64 | | matches | | |
| 98.21% | Filter word density | Target: ≤3% sentences with filter/hedge words | | filterCount | 0 | | hedgeCount | 2 | | narrationSentences | 64 | | filterMatches | (empty) | | hedgeMatches | | |
| 100.00% | Gibberish response detection | Target: ≤1% gibberish-like sentences (hard fail if a sentence exceeds 800 words) | | analyzedSentences | 75 | | gibberishSentences | 0 | | adjustedGibberishSentences | 0 | | longSentenceCount | 0 | | runOnParagraphCount | 0 | | giantParagraphCount | 0 | | wordSaladCount | 0 | | repetitionLoopCount | 0 | | controlTokenCount | 0 | | maxSentenceWordsSeen | 34 | | ratio | 0 | | matches | (empty) | |
| 100.00% | Markdown formatting overuse | Target: ≤5% words in markdown formatting | | markdownSpans | 0 | | markdownWords | 0 | | totalWords | 1018 | | ratio | 0 | | matches | (empty) | |
| 100.00% | Missing dialogue indicators (quotation marks) | Target: ≤10% speech attributions without quotation marks | | totalAttributions | 5 | | unquotedAttributions | 0 | | matches | (empty) | |
| 63.90% | Name drop frequency | Target: ≤1.0 per-name mentions per 100 words | | totalMentions | 39 | | wordCount | 813 | | uniqueNames | 16 | | maxNameDensity | 1.72 | | worstName | "Quinn" | | maxWindowNameDensity | 3 | | worstWindowName | "Quinn" | | discoveredNames | | Harlow | 1 | | Quinn | 14 | | Soho | 1 | | Morris | 4 | | Square | 1 | | Mile | 1 | | Nest | 2 | | Raven | 1 | | Silas | 2 | | Detective | 2 | | Tommy | 5 | | Despite | 1 | | Understanding | 1 | | Tube | 1 | | Camden | 1 | | London | 1 |
| | persons | | 0 | "Harlow" | | 1 | "Quinn" | | 2 | "Morris" | | 3 | "Raven" | | 4 | "Silas" | | 5 | "Detective" | | 6 | "Tommy" | | 7 | "Understanding" |
| | places | | 0 | "Soho" | | 1 | "Square" | | 2 | "Nest" | | 3 | "London" |
| | globalScore | 0.639 | | windowScore | 0.667 | |
| 100.00% | Narrator intent-glossing | Target: ≤2% narration sentences with intent-glossing patterns | | analyzedSentences | 57 | | glossingSentenceCount | 0 | | matches | (empty) | |
| 100.00% | "Not X but Y" pattern overuse | Target: ≤1 "not X but Y" per 1000 words | | totalMatches | 0 | | per1kWords | 0 | | wordCount | 1018 | | matches | (empty) | |
| 100.00% | Overuse of "that" (subordinate clause padding) | Target: ≤2% sentences with "that" clauses | | thatCount | 1 | | totalSentences | 75 | | matches | | |
| 100.00% | Paragraph length variance | Target: CV ≥0.5 for paragraph word counts | | totalParagraphs | 25 | | mean | 40.72 | | std | 24.14 | | cv | 0.593 | | sampleLengths | | 0 | 53 | | 1 | 73 | | 2 | 101 | | 3 | 82 | | 4 | 72 | | 5 | 74 | | 6 | 46 | | 7 | 30 | | 8 | 60 | | 9 | 49 | | 10 | 22 | | 11 | 21 | | 12 | 21 | | 13 | 33 | | 14 | 25 | | 15 | 12 | | 16 | 43 | | 17 | 40 | | 18 | 15 | | 19 | 21 | | 20 | 29 | | 21 | 5 | | 22 | 40 | | 23 | 35 | | 24 | 16 |
| |
| 94.30% | Passive voice overuse | Target: ≤2% passive sentences | | passiveCount | 2 | | totalSentences | 64 | | matches | | 0 | "was packed" | | 1 | "was revoked" |
| |
| 0.00% | Past progressive (was/were + -ing) overuse | Target: ≤2% past progressive verbs | | pastProgressiveCount | 4 | | totalVerbs | 130 | | matches | | 0 | "was going" | | 1 | "was going" | | 2 | "was dealing" | | 3 | "was going" |
| |
| 0.00% | Em-dash & semicolon overuse | Target: ≤2% sentences with em-dashes/semicolons | | emDashCount | 6 | | semicolonCount | 3 | | flaggedSentences | 8 | | totalSentences | 75 | | ratio | 0.107 | | matches | | 0 | "She had to catch this man; he was her only lead in a case that had gone cold two years ago." | | 1 | "She knew he was going to the Nest; the city's supernatural underworld wasn't infinite." | | 2 | "Another neon sign flashed by—'The Raven's Nest'—its distinctive green glow announced her next steps." | | 3 | "The place was packed with an eclectic mix of patrons—rough-looking men and women, their eyes sharp and alert, as if ready for a fight at any moment." | | 4 | "\"Same reason your Saint Christopher's dangling round your neck; we all want something,\" Quinn replied, her eyes still searching for her target." | | 5 | "A bone—a token to enter an underground market in an old Tube station beneath Camden." | | 6 | "The suspect had been boasting about a big score—was this it?" | | 7 | "Silas—rumoured to be a demonic figure who controlled much of the supernatural underworld in London." |
| |
| 100.00% | Purple prose (modifier overload) | Target: <4% adverbs, <2% -ly adverbs, no adj stacking | | wordCount | 416 | | adjectiveStacks | 0 | | stackExamples | (empty) | | adverbCount | 11 | | adverbRatio | 0.026442307692307692 | | lyAdverbCount | 3 | | lyAdverbRatio | 0.007211538461538462 | |
| 100.00% | Repeated phrase echo | Target: ≤20% sentences with echoes (window: 2) | | totalSentences | 75 | | echoCount | 0 | | echoWords | (empty) | |
| 100.00% | Sentence length variance | Target: CV ≥0.4 for sentence word counts | | totalSentences | 75 | | mean | 13.57 | | std | 6.64 | | cv | 0.489 | | sampleLengths | | 0 | 17 | | 1 | 16 | | 2 | 20 | | 3 | 15 | | 4 | 16 | | 5 | 21 | | 6 | 21 | | 7 | 14 | | 8 | 16 | | 9 | 21 | | 10 | 16 | | 11 | 22 | | 12 | 8 | | 13 | 4 | | 14 | 13 | | 15 | 18 | | 16 | 14 | | 17 | 18 | | 18 | 19 | | 19 | 14 | | 20 | 7 | | 21 | 20 | | 22 | 31 | | 23 | 10 | | 24 | 5 | | 25 | 11 | | 26 | 16 | | 27 | 7 | | 28 | 13 | | 29 | 12 | | 30 | 10 | | 31 | 10 | | 32 | 15 | | 33 | 11 | | 34 | 11 | | 35 | 19 | | 36 | 15 | | 37 | 12 | | 38 | 6 | | 39 | 27 | | 40 | 18 | | 41 | 14 | | 42 | 17 | | 43 | 12 | | 44 | 10 | | 45 | 7 | | 46 | 14 | | 47 | 2 | | 48 | 19 | | 49 | 22 |
| |
| 77.33% | Sentence opener variety | Target: ≥60% unique sentence openers | | consecutiveRepeats | 2 | | diversityRatio | 0.48 | | totalSentences | 75 | | uniqueOpeners | 36 | |
| 52.91% | Adverb-first sentence starts | Target: ≥3% sentences starting with an adverb | | adverbCount | 1 | | totalSentences | 63 | | matches | | 0 | "Then, with a final glance" |
| | ratio | 0.016 | |
| 100.00% | Pronoun-first sentence starts | Target: ≤30% sentences starting with a pronoun | | pronounCount | 17 | | totalSentences | 63 | | matches | | 0 | "She had to catch this" | | 1 | "It had left her clinging" | | 2 | "She had dealt with beings" | | 3 | "He was just a common" | | 4 | "They'd started chasing him across" | | 5 | "She knew he was going" | | 6 | "She is older than this" | | 7 | "Her gut told her it" | | 8 | "She had no backup now" | | 9 | "She could make out the" | | 10 | "She took a breath, steeling" | | 11 | "It was Tommy, an ex-paramedic" | | 12 | "He nodded towards an old" | | 13 | "It made sense, and it" | | 14 | "It was on its first" | | 15 | "He didn't plan to cash" | | 16 | "She turned her gaze back" |
| | ratio | 0.27 | |
| 71.11% | Subject-first sentence starts | Target: ≤72% sentences starting with a subject | | subjectCount | 49 | | totalSentences | 63 | | matches | | 0 | "The heavy rain blurred the" | | 1 | "Detective Harlow Quinn sprinted through" | | 2 | "The suspect's dark coat flapped" | | 3 | "Quinn's heart pounded in her" | | 4 | "She had to catch this" | | 5 | "It had left her clinging" | | 6 | "The suspect darted across an" | | 7 | "Quinn's worn leather boots splashed" | | 8 | "She had dealt with beings" | | 9 | "He was just a common" | | 10 | "This one made Morris stir" | | 11 | "Something didn't add up." | | 12 | "They'd started chasing him across" | | 13 | "Quinn weaved through the maze" | | 14 | "She knew he was going" | | 15 | "She is older than this" | | 16 | "Another neon sign flashed by—'The" | | 17 | "The suspect was going to" | | 18 | "Her gut told her it" | | 19 | "She had no backup now" |
| | ratio | 0.778 | |
| 100.00% | Subordinate conjunction sentence starts | Target: ≥2% sentences starting with a subordinating conjunction | | subConjCount | 3 | | totalSentences | 63 | | matches | | 0 | "While heading into the Nest" | | 1 | "If this guy was dealing" | | 2 | "If Quinn's gut was right," |
| | ratio | 0.048 | |
| 95.24% | Technical jargon density | Target: ≤6% sentences with technical-jargon patterns | | analyzedSentences | 45 | | technicalSentenceCount | 3 | | matches | | 0 | "Quinn weaved through the maze of trash bins and abandoned shopping carts, her mind racing with possible outcomes." | | 1 | "The place was packed with an eclectic mix of patrons—rough-looking men and women, their eyes sharp and alert, as if ready for a fight at any moment." | | 2 | "Silas—rumoured to be a demonic figure who controlled much of the supernatural underworld in London." |
| |
| 0.00% | Useless dialogue additions | Target: ≤5% dialogue tags with trailing filler fragments | | totalTags | 4 | | uselessAdditionCount | 2 | | matches | | 0 | "Tommy asked, his voice raising above the music" | | 1 | "Quinn replied, her eyes still searching for her target" |
| |
| 73.08% | Dialogue tag variety (said vs. fancy) | Target: ≤10% fancy dialogue tags | | totalTags | 3 | | fancyCount | 1 | | fancyTags | | 0 | "something he had (have)" |
| | dialogueSentences | 13 | | tagDensity | 0.231 | | leniency | 0.462 | | rawRatio | 0.333 | | effectiveRatio | 0.154 | |