What does the quality benchmark mean?

The quality benchmark is the Content Quality Score a draft needs before it is called ready — 7.0 for this X short-form rubric (benchmarks range 7.0–8.0 by industry). Scoring is rule-based and explainable: three samples with the median kept, filed with the run date, and every dimension shows why it scored as it did.

Can I run my own draft through the Analyzer?

Yes — paste any article or post on the Analyzer page. First scores are free with no card; signing up also includes 10 free article generations.

Real output · Real draft · Real score

Score your AI draft. See exactly what reads AI. Fix it.

Q: Is this a real report?

Yes. Every number on this page is a recorded run kept in the product's repo: the 6.2 draft and 7.6 final come from x_posts/_x_compare_v1_score.json and _x_compare_v35_scores.json, dated September 2026. No score here is invented.

Q: Does ForgeRank AI detect AI?

No. ForgeRank AI is a writing-quality gate, not an AI detector. It reports transparent patterns (sentences that read as AI) with concrete fixes — never fake percentages or '100% human' claims.

Paste any article or post — ForgeRank AI scores it on 4 transparent dimensions and lists the exact sentences that carry AI-sounding phrasing, each with a concrete fix. This is a real report from our Analyzer (real draft, real scores).

✓ Not a "probability" detector — transparent signals, no fake percentages · no “100% human” claims

The final draft — scored by our Analyzer

real X post · real run

7.6/10

✓ Ready (benchmark 7.0) PRO review 9/10 AI-tone: no signals found

Platform: X (Twitter) · short-form rubric
Character count: 241 / 280

Same fix, two reports: AI: “We leverage innovative solutions to optimize workflows.” Human: “We cut reporting from 3 days to 20 minutes.” AI writes what sounds smart. Humans write what actually happened. Checkable always beats impressive.

Verified metrics from this run

4-dimension diagnostic 7.5/10 · AI-tone lint 0 hits · CLEAN · Display standard PASS · 241/280 chars

Before → after: what the score actually caught

two real drafts, same topic — the fix came from our reviewer's flagged issues

BEFORE · DRAFT

Version 1 — first draft

6.2 / 7.0 line

Nearly ready · PRO 8/10 · not clean · 4-dim 7.5

4-dimension diagnostic (real per-dimension scores):

Real information gain7.0

Specificity density8.0

Source compliance8.0

Argument & mechanism depth7.0

Same sentence, two versions. AI: “We leverage innovative solutions to optimize workflows.” Human: “I moved us off spreadsheets. Saved 6 hours a week.” Specifics win because they're checkable. Polish is just abstraction wearing a suit. What's the most AI-sounding sentence you've caught yourself writing?

✗

False equivalence. The two examples aren't the same task, so the comparison is rigged. “Same sentence, two versions.”

✗

Cause asserted, not shown. “Checkable beats impressive” is stated, not demonstrated.

✗

Aphorism ending. The closer performs the polish it criticizes. “Polish is just abstraction wearing a suit.”

Our reviewer said: make both examples the same workflow report, keep the human voice, and end on a checkable consequence instead of a slogan.

AFTER · FINAL

Version 5 — revised draft

7.6 / 7.0 line

Ready · PRO 9/10 · clean · AI-tone 0 hits

✓

Same task, fair comparison. One workflow change, two phrasings.

✓

Checkable outcome. “3 days → 20 minutes” is a concrete, verifiable result.

✓

Ending states the mechanism instead of performing a slogan.

When signals are found, you see the sentence — and the fix

real lint output on an internal signal-heavy test draft · real fix strings from our fixer · incl. new structure-level signals

Spoken-filler tick×2

“Here's the thing” · “The truth is”

Fix: cut the verbal tics and state the point plainly — the fact should open the sentence.

Buzzwords & padding connectors×1

“the truth is”

Fix: replace buzzwords and filler connectors (unlock, game-changer, furthermore) with plain, specific words.

Em-dash overuse×1

“—”

Fix: replace “—” with a comma or a new sentence; keep at most one em dash per sentence.

Known AI-typical pattern×6

“That's it. That's the whole strategy.”

Fix: reword in plain, specific language — say exactly what you mean instead of the template phrase.

Negative restructure (“not X — it's Y”)×1

“You don't need five hacks. You need one thing that works.”

Fix: say what it is — not what it isn't.

“Not X. Y.” staccato×2

“Not a trend. A habit.”

Fix: write it as a full sentence that states what it is.

Negative-listing buildup×1

“Not a hack. Not a gimmick. A real workflow.”

Fix: cut the repeated “Not a…” runway and state the positive directly.

Inanimate subject doing a human verb×1

“The strategy decides what ships next.”

Fix: name the person who acts instead of letting decisions/data/culture “decide, emerge, or tell us.”

The score archive — more real runs from the same gate

each kept in the repo with its date · nothing invented

Google long-form · editorial QC record 2026-09-08

How to Clean a Washing Machine the Right Way

8.2 / 10 · industry Home & Garden · pass line 7.5

Real editorial QC score for the published text (record: scripts/_blog_editorial_qc.json) — the same article that's live on this site. Open the article →

X · recorded run

"The 3 tells of fake AI writing"

8.8 / 10 · Ready · lint clean

Real short-form score from the same corpus as this page's before/after (record file: x_posts/_x_final.json).

IG + TikTok · one brief

"Spot AI content in three seconds"

9.2 / 9.2 · Ready

The same brief written for Instagram and TikTok, both scored (record: x_posts/_display_ig_tiktok_short_scores.json).

More reports are assembled from recorded runs as they happen — next up: a determinism check (the same brief scored repeatedly) and competitor benchmarking. If a run doesn't exist yet, we don't invent a number to fill the card.

Questions about this report

Is this a real report?

Yes. The 6.2 draft and the 7.6 final are recorded runs kept in the repo — x_posts/_x_compare_v1_score.json and _x_compare_v35_scores.json, September 2026. The footer lists the exact file for every number on this page.

What does "Ready (benchmark 7.0)" mean?

7.0 is the quality benchmark for this X short-form rubric — the Content Quality Score a piece needs to be called ready (benchmarks run 7.0–8.0 depending on industry). Scoring is rule-based and explainable: three samples with the median kept, filed with the run date, and every dimension shows why it scored as it did.

Does ForgeRank AI detect AI?

No — it's a writing-quality gate, not an AI detector. It reports transparent patterns (the exact sentences that read as AI) with a concrete fix for each. No fake percentages, no "100% human" claims.

Can I run my own draft through this?

Yes — paste any article or post in the Analyzer. First scores are free with no card; signing up also includes 10 free article generations.

Run your own draft through the same analyzer

Paste any article or post. Get the 4-dimension score, the flagged sentences, and a concrete how-to-fix in seconds. First scores are free — no card.

Try ForgeRank AI free10 free articles · no card · first scores free

Credit packs from $4.99 one-time — 75 credits on first purchase (≈ $0.0666/piece) · no subscription · credits never expire