Skip to main content
theStacc shows an SEO score from 0 to 100 in a post’s Analysis panel, with category checks and suggestions. Use those suggestions to find structural gaps and review the writing. The score is an internal assessment, not a Google ranking score or a guarantee that the article is accurate.
Product checks, not search requirements: The weights and thresholds below describe theStacc’s documented scoring rubric. Inspect the checks shown for your post, and evaluate whether each suggested edit helps the reader. Meeting a numerical target does not establish a ranking outcome.

Two scores blended into one

The headline combines two assessments with different limits:

Mechanical and semantic assessment limits

AssessmentWhat it examinesWhat it does not establish
Mechanical scoreRules-based checks of the article’s structure, such as headings, word count, links, and alt textWhether every factual claim is correct or the advice fits your business
Semantic scoreAn AI assessment of topic coverage, search intent, and helpfulness during generationIndependent expert approval, verified research, or performance in search results
The two assessments combine as follows: Overall = (0.4 × mechanical) + (0.6 × semantic) Those weights belong to theStacc’s rubric. They do not describe the relative importance of factors in Google’s ranking systems. When a semantic score is unavailable, the fallback uses the mechanical score alone. The Analysis panel shows the headline and a Mechanics N - Semantic M breakdown. Check which assessments are present before interpreting a change in the headline.

Rating bands

The rating labels summarize the numerical result:
  • 90 or above: Excellent
  • 80 to 89: Good
  • 70 to 79: Needs Work
  • 60 to 69: Poor
  • Below 60: Critical
These are product labels. An Excellent score does not replace a review of the claims, examples, audience fit, or publication permissions. A lower score identifies suggestions to investigate; it does not prove that a useful, focused article is unsuitable for publication.

What the mechanical score checks

The mechanical rubric has nine positive categories with a combined maximum of 100 points, plus a penalty of up to 15 points. The category values help you understand the analysis; they are not a checklist of Google’s required quantities.

Keyword placement (10 points)

Checks the target keyword in the title, H1, and URL slug. The matching allows words in a different order. For example, “Warehouse Skylight Installation Cost” and “Skylight Installation Cost for Warehouse” can match the same target in the product’s check. Use wording that describes the page naturally. Product matching does not demonstrate that two phrases have identical search intent or receive identical rankings.

Content length and depth (15 points)

The length check favors roughly 1,000 to 4,000 words, along with subheading coverage and paragraph size. It also favors five or more subheadings. These are internal scoring preferences, not universal article requirements. Google’s people-first content guidance says Google has no preferred word count. Review missing explanations, repeated sections, and the reader’s task before expanding a draft. See Word Count & Targets for a decision framework.

Readability (5 points)

Uses reading-ease and sentence-length checks for English text. Its sentence-length preference is roughly 10 to 25 words. Preserve technical accuracy and natural rhythm rather than forcing every sentence into the same range. These checks are English-specific. See Writing in other languages for how the documentation describes that exception.

Heading structure (10 points)

Checks the heading hierarchy and spacing of sections. Use one page-level H1, H2s for main sections, and H3s for subsections. The scoring guidance describes a subheading roughly every couple of hundred words; add one when the topic needs a new section, rather than interrupting a connected explanation. The generation guide describes body HTML that starts at H2 because the CMS supplies the title. Review the complete rendered page before adding another H1 to the body. The link check rewards three or more internal links, two or more external links, and descriptive anchors. It skips the internal-link quantity check when the project has no other published blogs. Choose links that help the reader navigate or verify a claim. A link count alone does not establish relevance, source quality, or a working destination. Do not add unrelated links to reach the target.

Images and media (5 points)

Checks for two or more images and alt text. These quantities are scoring preferences. Use media when it explains the subject, and make the text alternative describe the image’s purpose. Adding decorative images solely for points can make the article harder to use.

Technical meta (5 points)

The meta checks include a meta description of at least 50 characters and a title between 30 and 65 characters. These ranges help identify fields to review; they cannot guarantee that a search result will display the full text. Write a clear title and description that match the page.

Entity and topic coverage (25 points)

Uses text signals such as vocabulary variety and whether the body addresses topics promised by its headings. Treat this category as a proxy for coverage. It cannot by itself verify that a named product, regulation, statistic, or example is accurate. Read the explanation and check its evidence.

Search-intent match (15 points)

Compares the structure with the selected blog type. A comparison may need a table or useful tradeoffs; a how-to needs a workable process. The next section describes the documented format checks. They should support the reader’s task rather than dictate unnecessary sections.

The spam and AI-pattern penalty (up to -15)

This internal penalty can subtract up to 15 points for patterns the product flags:
  • Generic opening phrases, such as “In today’s fast-paced world.”
  • Keyword repetition, with roughly 5% density used as an internal flag.
  • Heavy dash use in the prose.
  • Repeated filler transitions or paragraph openings.
These are review signals. A percentage is not Google’s definition of keyword stuffing, and punctuation alone does not establish that content is spam. Google’s keyword-stuffing policy concerns manipulative, unnatural repetition. Edit the passage for meaning and readability, including any valid quotations or technical terms, instead of assuming every flagged word must disappear. Review generated drafts as well as manually edited ones. This guide does not establish that AI-generated content always avoids penalties or that an internal flag predicts a search-engine action.

Search-intent expectations by blog type

The format checks look for these structures:
  • How-to / Tutorial: numbered step headings, with four or more steps favored by the rubric.
  • Listicle: numbered item headings, with five or more favored.
  • Ultimate guide: broad coverage, with six or more H2s and 1,500 or more words favored.
  • Comparison: a comparison table, pros and cons, or clear versus headings.
  • Case study: a problem, action, and outcome supported by concrete evidence.
  • Beginner’s guide: a simple process and useful questions answered.
  • Problem-solution: clear descriptions of the problem and proposed solution.
  • News and trends: a relevant time anchor and supporting facts.
The fallback scores this category neutrally when no blog type is set. See Blog Type Options to choose the format that fits the article. Do not invent numbers for a case study, add irrelevant FAQs, or change a publication date to satisfy a format check. Keep the source’s date and limitations visible, and use only the sections the topic needs.

Writing in other languages

The documentation describes skipping English-only readability checks for predominantly non-English posts and awards credit for the skipped checks. This means a passing readability result may indicate that the test was not applied. It does not certify grammar or clarity in the article’s language. The documentation also describes Unicode-aware topic-coverage handling. Have a fluent reviewer assess language-specific terminology, examples, and tone; a text-processing check cannot supply that review.

How the AI rates itself (the semantic score)

The semantic assessment considers topical depth, search-intent match, and helpfulness. Keyword placement is handled by the mechanical checks rather than repeated in this assessment. The generation rubric describes an intended range of 70 to 85, with 90 or above reserved for a stronger assessment of coverage. These are instructions for the product’s assessment, not observed benchmarks showing that an article outperforms current search results. The AI’s own rating is not independent verification of the writing it generated.

Where the score shows up, and how it updates

Open the post’s Analysis panel and inspect:
  1. The headline score and rating.
  2. The Mechanics N - Semantic M breakdown, where available.
  3. The category scores and expandable suggestions.
Read a suggestion before acting on it. A missing source may need research; a missing link may need a relevant destination; an overlong section may need a narrower scope. The goal is to resolve the specific gap, not merely move the headline number.

Recalculating after you edit

Recalculation distinguishes edited structure from the original generated assessment:

What recalculation updates

Part of the analysisWhat recalculation does
Mechanical scoreRuns the rules-based checks against the edited content
Semantic scoreKeeps the assessment returned during the original generation
Headline scoreCombines the available assessments again
A new headline score can therefore include a semantic assessment of an earlier draft. Re-read any substantially rewritten sections yourself. Recalculation does not establish that added facts were verified or that the AI has reassessed every edit.

Auto-regeneration of low-scoring posts

Earlier scoring documentation describes retries for posts below a quality threshold. Check the current generation status and suggestions rather than assuming that every low-scoring draft is automatically blocked from publishing. See Blog Generation for the documented generation workflow. Scoring and publishing approval are different controls. In Draft for me, you review the draft before publication; Full autopilot publishes without that human approval step. Confirm the selected mode and review responsibilities, especially for sensitive claims.