Review scores
Lyra reviews every draft before you see it and gives it a score out of 100. The score reflects editorial quality, SEO, answer-engine readiness, and accuracy. A post does not reach Ready until it clears the bar with no outstanding issues.
This page explains what the score measures and how to read the reviewer panel on a post.
The verdict
The reviewer panel shows a headline score and a verdict:
| Score | Verdict |
|---|---|
| 85 and above | Pass. Ready to merge. |
| 75 to 84 | Needs minor work. |
| Below 75 | Needs work. |
Lyra keeps iterating until the post passes. If she cannot clear the bar within the maximum number of iterations, she opens the pull request as a draft, marks the post Needs human, and lets you decide what's next. See Reviewing and merging.
Scoring categories
Lyra scores each draft across eight dimensions. Every category contributes to the overall score.
| Category | What it measures |
|---|---|
| Content | Depth, structure, and usefulness of the writing |
| SEO | On-page optimization for search |
| AEO | Answer-engine optimization, so AI assistants can cite the post |
| E-E-A-T | Experience, expertise, authoritativeness, and trust |
| Accuracy | Whether claims and facts hold up |
| Readability | Clarity and flow for a human reader |
| Linking | Internal and external links, including your tracked links |
| Voice | How closely the draft matches your blog's voice |
Issues
Alongside the score, Lyra lists the specific issues she found, each with a severity:
- High: a blocker, such as a broken link, a contradicted claim, or a broken rule from your writing guide.
- Medium: a meaningful problem worth fixing before merge.
- Low: a minor improvement.
She posts these as resolvable line comments on the pull request and resolves them as she fixes them.
Iteration
Lyra re-reviews after every push. Each pass through the review-and-fix loop is one iteration, shown as Iter in the Blogs table. When she reviews a revised draft, she first checks which of her earlier notes it already fixed, so she does not raise the same point twice. A small note no longer blocks the final writing pass, so fewer posts come back to you over minor things.
What the review checks
Beyond the eight scores, the review reads the draft against three things:
- Your writing guide, rule by rule. Every rule in your guide is binding. Breaking one counts as a real problem, not a passing note, and the review quotes the rule and the exact line that broke it.
- The post's search plan. Each subtopic, question, and angle Lyra planned from the top search results is checked against the draft one by one, so the same draft gets the same verdict every time.
- Your site's subject. A post that drifted away from your content focus is caught before it reaches you.
The review also checks up front whether the post answers its reader's question from the first paragraph, whether the title promises only what the post delivers, and whether it shows first-hand experience, such as a worked example or a real measurement.
Fact verification
Every factual claim is checked against the page it cites, not only that the link loads. The fact check runs once, on a post that has cleared review, using a stronger model than the one that wrote the post, so it is a real second opinion.
For each claim, Lyra reads the cited page and decides:
| Finding | What happens |
|---|---|
| Confirmed | The claim stays as written |
| Wrong figure | The figure is corrected to what the source says |
| Contradicted | The source says something else. The sentence comes out if the post's point holds without it |
| Not confirmed | Lyra could not find the claim on the page, or could not read the page. The claim stays and is reported to you |
A claim is only removed when Lyra has read the page it cites and found that the page says otherwise. Not finding something is not the same as finding it wrong: a page behind a sign-in, one that draws itself in your browser, or one rewritten since can all hide a true claim. Figures taken from a table are checked against the numbers in that table.
If a claim central to the post's argument cannot be confirmed, Lyra leaves it for your call instead of publishing it or quietly removing it. See Reviewing and merging.
Every pull request ends with a fact verification summary: how many claims were checked and confirmed, and what was corrected, dropped, or left for you.
Price verification
If you set a pricing page in Settings > Website, Lyra also verifies any prices the post cites against that page and requires a "current as of post date" note. Price verification only runs when a pricing-page URL is set. See Configure your website.
Where you see scores
- Blogs shows the score for each post in the table.
- Overview shows your Avg review score across recent posts.
- Blog detail shows the full reviewer panel: the headline score, per-category scores, and the list of issues.
What's next
- The pipeline: Where review sits in the flow.
- Reviewing and merging: Read the panel and merge.
- Configure your website: Set tracked links and your pricing page.