The Content Score, criterion by criterion.

Eight weighted checks that predict whether an AI answer engine will quote your article. Every point is traceable — and the 20-point answer block is the one people break when they edit.

The Content Score is the gauge in the right-hand rail of the editor, and it answers exactly one question: how likely is an AI answer engine to quote this article? It is not a Google ranking prediction and it is not a grammar grade. It's a citability rubric — eight things that make a page quotable, each worth a fixed number of points.

What the number means

The score runs 0–100 and is the sum of eight weighted criteria. There is no AI involved and no magic: it's a set of checks run against your article's text and its structured data. That's deliberate — an opaque score you can't act on is just a number to feel bad about.

Two things follow from that:

It recomputes as you type. The rail rescores your body a moment after you stop typing. It costs nothing — no credits, no AI call.
Every point is traceable. Open Score breakdown under the gauge and you get all eight rows with earned/possible and a one-line hint on each one you haven't maxed out.

When you first open a finished article, the gauge shows the score your article was saved with. The moment you edit the body it switches to the live recompute and starts showing a delta (▲ +6 since last save) so you can see whether your edit helped or hurt.

The four bands

The pill above the gauge tells you which band you're in.

Band	Score	What it says
Elite	85–100	Top of the field
Strong	70–84	Beating most competitors
Competitive	50–69	Closing in on the field
Needs work	0–49	Below the target band

Under the gauge you'll see Top-10 SERP avg ~68. That's a fixed reference point, not a live measurement of your keyword's actual top ten — it's the benchmark the rubric is calibrated against. Treat it as the bar to clear, not as data about your competitors.

// AIM FOR THE BAND, NOT FOR 100

The target is 70–90. Chasing a perfect 100 means forcing a HowTo section into an article that isn't a how-to, and stuffing in quotes nobody said. A Strong article that reads like a human wrote it beats an Elite one that reads like it was written for a checklist.

The eight criteria

Criterion	Points	Scoring
40–60 word answer block at the top	20	All or nothing
External citations	15	Graded
FAQPage schema	15	All or nothing
An attributed quote	10	All or nothing
HowTo schema	10	All or nothing
Author schema	10	All or nothing
Entity coverage	10	Graded
Verifiable claims	10	Graded

1 · The answer block — 20 points

The single biggest lever, and the one people break most often when they edit. It's the paragraph an AI Overview lifts verbatim.

Full marks: your first H2 section contains one paragraph of roughly 40–60 words that answers the article's question directly. (The check accepts 35–70 words, so you have a little room.)
Loses it: you split that paragraph in two, pad it past 70 words, trim it below 35, or move it under the second H2. It's 20 points, all at once — this is almost always why a score drops after an edit.

2 · External citations — 15 points

Links out to sources that aren't you. A link to your own domain doesn't count.

3 or more external links — 15 points
2 links — 10 points
1 link — 5 points
None — 0

Any link to a domain other than your own counts, but authority matters for the reader even where the score is indifferent. The linking strategy card tiers what you've got and suggests more.

3 · FAQPage schema — 15 points

Earned when the article's structured data carries a FAQPage block. It's generated for you at the end of the write, as long as the FAQ setting is on in the sidebar.

Full marks: the FAQ block exists. Check it in the FAQ schema (JSON-LD) card under the article.
Loses it: you turned the FAQ setting off before generating. Turn it back on and use Regenerate on the FAQ card (1 credit).

4 · An attributed quote — 10 points

One direct quote with a name on it. The check looks for a quoted passage of roughly 20–400 characters followed by an attribution:

both of these count// em-dash attribution
"We stopped chasing rankings and started answering questions." — Sarah Mills, Head of Growth

// parenthesised attribution
"Most local firms never audit their own site." (James Okafor)

Full marks: at least one quote in that shape, from a customer, your founder, or a named expert.
Loses it: a quote with no name after it, or an attribution that doesn't start with a capitalised name. A pull-quote from “a client” earns nothing.

5 · HowTo schema — 10 points

Only relevant when the article is genuinely a step-by-step. It's emitted when one of your outline's sections carries the HowTo schema hint and that section has sub-headings for the individual steps.

Full marks: tag a procedural section as HowTo in the outline, and give it H3s — one per step.
Loses it: the article isn't procedural, or the HowTo section has no sub-headings.

If your article isn't a how-to, accept the missing 10 and aim for 90 instead of 100. Forcing steps into an explainer makes it worse, not more citable.

6 · Author schema — 10 points

Earned when the structured data carries an author Person block. It is built for you from your brand's name, website and About text, so in practice this one lands on its own — it only goes missing when the article has no structured data at all (which also costs you the FAQ's 15).

Fill in the About field on your brand profile anyway: it's what puts a real description in the author block rather than a bare name, and author credentials are one of the strongest signals an AI assistant reads.

7 · Entity coverage — 10 points

Unique named entities — people, organisations, products, places, tools and named concepts — found in the finished article. Graded:

20+ entities — 10 points
15–19 — 7 points
10–14 — 4 points
5–9 — 2 points
Under 5 — 0

The number comes from the Entities card under the article, which also suggests entities you could reasonably add. Naming the actual tools, standards, cities and companies your topic involves is the cheapest way to move this.

8 · Verifiable claims — 10 points

The subtlest one. We look at every paragraph containing a statistic — a percentage, a figure with a thousands separator, a multiple like 3x, or a phrase like “40 percent” — and check whether that same paragraph also contains a link. The score is the proportion that do.

Half your stats sourced or better — 10 points
30%+ — 6 points
10%+ — 3 points
Below that — 0

// AN ARTICLE WITH NO NUMBERS SCORES ZERO HERE

If there isn't a single statistic in the body, the ratio is zero and so are the points. That's not a bug — an article with no verifiable claims has nothing for an AI assistant to cite. Add two or three real figures with a source link beside each.

Raising your score, in order

Work down this list. It's ordered by points-per-minute of effort.

Fix the answer block (up to 20)

Open the first H2. Make sure there is exactly one paragraph of about 50 words that answers the title's question outright — no throat-clearing, no “in this article we'll explore”. If the breakdown shows this row at 0, this is your whole afternoon's work in one paragraph.

Get to three external links (up to 15)

Open the Linking strategy card. It lists what's already in your body and suggests authoritative sources to add. Paste two or three in where they support a claim.

Source your statistics (up to 10)

Find every number in the article and put a link in the same paragraph. This and step 2 are the same edit done twice — do them together.

Add one quote with a name (10)

A line from your founder or a customer, with their name after an em-dash. Ten points for one sentence.

Name more entities (up to 10)

Open the Entities card and read the “suggested to add” row. Work the relevant ones into the body naturally. Getting from 12 to 20 entities is worth 6 points.

Check the FAQ block exists (15)

If the FAQ schema row reads 0, the FAQ setting was off when you generated. Turn it on and hit Regenerate on the FAQ card.

The Re-optimize button

Under the gauge you'll sometimes see an orange panel offering to Re-optimize to 80+. It appears only when the article's saved score is below 80. If your article is already in the band, the button doesn't exist — there's no manual step in the happy path, because new articles are optimised during generation.

It runs the same optimisation pass the writer runs, weaving in missing SERP terms and topping up thin sections, then re-scores and reloads the page with the improved article. It's free — no credits.

// IT REWRITES YOUR BODY

Re-optimize edits the article in place. If you've made careful manual edits, save a version first (⋮ → Version history → Save current as version) so you can get back.

Things that surprise people

The score moved and you didn't touch the text. The entity count arrives from the Entities card a moment after the editor loads, and the score updates to match. That's the two numbers agreeing, not the score drifting.
The score dropped after a re-tune. Re-tuning a passage rewrites it — if that passage was your answer block, or held your only sourced statistic, you lose those points. Check the breakdown before you accept a rewrite.
Scoring is free and local. Nothing is billed for computing the score, no matter how many times it recomputes.
The saved score and the live score can differ. The gauge shows the saved one until your first edit. That's on purpose — a freshly opened article should never flash a number lower than what the pipeline actually earned.

Still stuck?

Contact support