Why reviews drift
Two editors, same draft, different verdicts.
Without a rubric, every review is a taste test. One editor flags structure, another flags sourcing, and the writer learns that approval depends on who picks up the piece. Feedback arrives as prose, so nothing is comparable across drafts, and a manager cannot tell whether quality is improving or the reviewer just changed.
A scorecard fixes this by doing three things prose feedback cannot: it scores the same categories every time, it separates blockers from polish, and it produces a number you can track by writer and by month. The rubric below is the one-page version.
The rubric
Five categories, five points each
Score every category 0-5. Read the piece once as a reader, then score in one pass. The category descriptions are the standard; if a score needs defending, quote the draft, not your impression of it.
Required sections present, assigned angle held, internal links placed with assigned anchors, CTA matches the target page. A 5 needs zero missed requirements.
Every stat, quote, and claim traces to an approved source. Score 2 or below when any specific claim is unsourced; unverifiable means unpublishable.
Piece matches the search intent in the brief, target keyword used naturally in title and headings, no stuffing, no drift into a different query.
Reads like the client, not like a template. Generic filler ("in today's digital world") and sections that could apply to any company cap this at 2.
Scannable headings, tight paragraphs, logical order, intro that names the reader's problem. Judge as the reader who searched, not as an editor.
Then apply the decision line:
SCORE DECISION
21-25 Approve. Log any polish notes for next time.
15-20 Minor revision. List the exact items; one round.
Below 15 Major revision. Tag the reason (brief gap,
writer gap, or standard gap) before reassigning.
BLOCKERS (automatic major revision at any score):
- A fabricated or unverifiable source
- A missed mandatory brief requirement
- Plagiarism or undisclosed AI-generated sections
- Wrong search intent for the assigned queryHow to use it
Make the score do the arguing
Share the rubric with writers before the first review. A scorecard the writer has never seen is a trap, not a standard. Most writers self-correct against a visible rubric, which is the cheapest QA you will ever run.
Score the piece, not the person. The score judges one draft against one brief. Patterns across drafts are a coaching conversation, and by then you have numbers instead of impressions to have it with.
Separate blockers from points. A beautifully written piece with one invented source is not an 18. The blocker list exists so that the worst failure modes cannot be averaged away by good prose.
Return revisions with the category, the quote, and the fix. "Accuracy 2/5: the pricing claim in section 3 has no source; use the approved pricing page" is feedback a writer can execute in one round. "Make it more credible" is not.
When this stops being enough
A scorecard measures one draft. A system remembers.
This rubric works well for one-off reviews and small volumes. It strains when you review more than about 10 pieces a month, when more than one reviewer needs to score the same way, and when scores need to connect to something: revision deadlines, writer history, payout decisions, and the question of which briefs keep producing low scores.
At that point the rubric should live inside a tracker, so every review is a logged record with a writer, a score, a status, and a deadline instead of a number in an editor's head.
Two next steps
The PDF version, and the tracker version
The free 25-Point Freelancer QA Rubric PDF is the deeper standalone version of this page: the full rubric with per-point definitions, a reviewer drift test for calibrating multiple editors, fair-use and blocker rules, and a completed example scorecard.
The Freelancer Ops System ($49) is the tracker version: the same QA rubric wired into a connected Notion workspace with submission tracking, revision logs with owners and deadlines, payout reconciliation, and a manager dashboard, so scores become visible patterns instead of one-off verdicts. It ships with example data pre-filled, includes free updates forever, and carries a 14-day fit guarantee.
Related resources
Fix the inputs the scorecard keeps flagging
Low brief-compliance scores usually mean the brief was the problem: start with the SEO content brief template. Repeated accuracy failures on AI-assisted drafts mean the rules were never set: see the AI use policy template for content teams.
FAQ
Common questions
Is content QA just proofreading?
No. Proofreading catches surface errors. Content QA judges whether the piece does its job: follows the brief, supports its claims, matches the search intent, and reads like the client. A typo-free draft can still fail QA badly.
What if two reviewers still score differently?
Run a drift test: both score the same two drafts independently, then compare per category. Gaps of 2+ points mean the category description is ambiguous; tighten the description rather than arguing the verdict. The free PDF includes this calibration exercise.
Should writers see their scores?
Yes, with the category breakdown and quotes. Hidden scores create fear without improvement. Visible scores with specific evidence turn QA into training you do not have to schedule.
Does this work for AI-assisted content?
Yes, and the blocker list is where AI risk concentrates: undisclosed AI-generated sections and fabricated sources are automatic major revisions regardless of the total score. Pair the scorecard with a disclosure rule so editors know what to check first.