Methodology / Change record
Scoring Model Version History
Only repository-backed history is published here. Unknown release details remain unknown.
Version policy
A scoring version records changes that can alter deterministic output. This includes a trigger, threshold, deduction, mode requirement, weight, tokenization rule, aggregation rule, severity boundary, or score band. Copy corrections that do not alter analyzer output do not create a new scoring version.
A version entry is published only when source, tests, fixtures, or a contemporaneous changelog retain enough evidence. Repository dates are not presented as public deployment dates.
SEO Blitz Scoring Model v2.1
| Model identifier | v2.1 |
|---|---|
| Model date in shipped source | October 2026 |
| Implementation commit | d7c9afbe60ee015e18d29087285df17ec7d90ca1 |
| Rules affected | STR-01, STR-02, STR-03, STR-04, CLR-02, RDB-01 to RDB-04. Rule IDs, weights, thresholds, deductions, and score bands are unchanged. |
| Previous behavior | Every ., !, or ? ended a sentence, including decimals, domains, e.g., and list numbers. Bullet lines counted as headings. Pasted HTML headings counted as zero. A line break before any capital letter started a paragraph. |
| New behavior | Sentences end only before whitespace or the end of the text, with abbreviation handling. Blank lines end sentences. Headings are Markdown # lines, whole-line bold, or pasted h1 to h6. Heading lines are excluded from sentence and paragraph analysis. Paragraphs break at blank lines, or at a sentence-ending line followed by a capital. |
| Reason | v2.0 split "The desk is 2.5 ft deep, e.g. for a laptop. Visit seoblitz.app today." into 6 sentences instead of 2, which distorted sentence-length rules. |
| Measured effect | All 163 retained Labs cases were re-run on v2.1. 162 are unchanged. One synthetic case (STR-02-negative, three unpunctuated paragraphs) moved from 85 to 84, because blank lines now end sentences. No target-rule result changed, so the 30 / 30 rule validation holds. All seven Example fixtures are identical. |
| Reproduction evidence | node scripts/labs-model-rerun.mjs --model v2.1 --commit d7c9afb... writes new result sets under data/labs/ with the suffix model-v2-1. v2.0 result sets run on the frozen engine data/labs/engines/analyzer-v2.0.js and still reproduce byte for byte. |
v2.1 is a parsing release. A score moves only where the text is segmented differently. Published v2.0 Labs studies remain the record of v2.0 behavior.
SEO Blitz Scoring Model v2.0
| Model identifier | v2.0 |
|---|---|
| Model date in shipped source | August 2026 |
| Repository implementation evidence | First tracked v2.0 implementation found in commit 4543f9025a31cdabe9ad5515f6fead9fbdc6675b, committed 2026-08-04 |
| Dimensions | Clarity, Completeness, Structure, Readability, Specificity, Repetition, Title alignment, Publication readiness |
| Modes | Blog, Landing, Product, Local, Help, Comparison |
| Overall calculation | Rounded mode-weighted sum of eight clamped dimension scores |
| Finding presentation | Evidence-linked findings with three prioritized revisions |
The retained source changelog says v2.0 added six modes, an optional brief, eight dimensions, evidence-linked findings, and top-three prioritization. It also records the move to a mode-weighted overall score and eight breakdown cards.
The commit date proves the first tracked repository implementation found. It does not prove when v2.0 was deployed publicly.
Historical inline v1.1 implementation
| Version identifier | v1.1 |
|---|---|
| Date retained in current changelog | May 2026 |
| Repository evidence | An inline analyzer exists in the parent of commit 4543f9025a31cdabe9ad5515f6fead9fbdc6675b |
| Verified behavior | Word, sentence, paragraph, title, and reading-time metrics; one combined score; five breakdown cards; repeated-term output; keyword cloud |
The current embedded changelog describes v1.1 as a single-score model with five breakdown categories. No retained evidence supports earlier public claims about FAQ validation or a threshold change relative to v1.0.
v1.0 is not reproducible from retained evidence
No reproducible v1.0 implementation, fixture, or verified changelog entry was found in the inspected repository history. SEO Blitz therefore does not publish a v1.0 feature list, thresholds, weights, model date, or release date.
Current evidence gaps
- The first tracked v2.0 repository implementation is known. Its external production deployment date is not verified.
- v2.1 shipped to production with commit
d7c9afbthrough the repository's Vercel deployment. The exact public deployment time is recorded by Vercel, not by this page. - The retained v1.1 inline engine has no dedicated frozen fixture in the current repository.
- No reproducible v1.0 engine was found.
Future change entries
Each future score-affecting change should identify its version, date, rule ID, previous behavior, new behavior, reason, expected effect, reproduction evidence, and implementation commit.
Read the current Methodology, inspect all v2.1 scoring rules, or review the model's limitations.