Project · History
aiseo-audit history: versions, releases, scoring, and baselines
In brief. This is the aiseo-audit history from the geoaudit prototype through each public version, sitemap audits, continuous integration, the Model Context Protocol server, and the evidence-tiered version 2 scoring model.
Maintained by Jeff Patterson and Agency Enterprise · Updated August 30, 2026
Why the project began: overview
The history of aiseo-audit records how an AI SEO tool replaced broad heuristics with evidence-tiered checks. Early advice mixed content practices, proposed standards, traditional SEO habits, and limited GEO research in one checklist. The first tool made those ideas testable by fetching a page and reporting observable signals.
According to the npm version archive [1]and the project's GitHub releases [2], the public changelog spans the original February 2026 prototype, the 1.x expansion, and the research-driven 2.0 rebuild.
That goal remains, but the standard for a trustworthy score changed. Repeatability was not enough. Each factor also needed a reason, a pipeline stage, and an explicit statement of the evidence limits. This means the version history records both new features and checks removed from the score.
Version timeline
A first open-source prototype
The project began as geoaudit, a command-line experiment that turned early GEO guidance into a page report. The first releases added the fetch, analysis, score, and report loop.
Source and release details ↗Published as aiseo-audit
The package took the aiseo-audit name and moved to npm. Early 1.x releases improved reports, configuration, continuous integration, and analysis while keeping one command as the main workflow.
Source and release details ↗From single pages to sitewide auditing
Version 1.3 improved robots.txt checks, entity detection, numeric claims, and image diagnostics. Version 1.4 added sitemap audits, shared site signals, host summaries, and stricter option checks.
Source and release details ↗A rapid scoring and NLP refinement cycle
A series of patch releases changed entity extraction, recommendations, scoring, documentation, and grades. The pace exposed a problem: a useful heuristic can turn into an unsupported assumption when no evidence record sits beside it.
Source and release details ↗Answerability, GitHub Actions, and Model Context Protocol
Version 1.5 added answerability checks, a GitHub Action for pull-request gates, and a Model Context Protocol server. Every interface used the same analyzer.
Source and release details ↗The final 1.x baseline
The 1.6 line prepared the Model Context Protocol Registry release and stabilized the last 1.x build. That build became the baseline for a review against ten newer peer-reviewed papers.
Source and release details ↗A research-driven rebuild
Version 2.0 split the citation pipeline into stages and added an evidence tier and paper citations to each factor. It also added structural alignment, target queries, page profiles, and evidence gates.
Source and release details ↗The current release
Version 2.0.1 is the current npm release in this history. The project now calls itself a research-informed heuristic audit. Major scoring changes require a migration and new baselines.
Source and release details ↗aiseo-audit history terms
geoaudit
geoaudit refers to the original command-line prototype released in February 2026.
aiseo-audit
aiseo-audit refers to the npm package name adopted for the 1.x releases.
Major version
A major version means that scoring or compatibility changed enough to require a new baseline.
Migration guide
A migration guide is defined as the record of changes needed between major versions.
Evidence tier
An evidence tier refers to the strength and scope assigned to an audit factor.
Diagnostic
A diagnostic is a type of visible observation excluded from the score.
What we got wrong in 1.x
Version 1 used the evidence and assumptions available at the time. A later review identified claims that were too confident or contradicted by controlled experiments.
Formatting was over-rewarded
Lists, tables, scannability, section length, and similar counts had little consistent causal effect when isolated. They remain diagnostics, but no longer add points by their mere presence.
Repetition was treated too generously
Newer retrieval studies showed added keyword density hurt in tested settings. Entity Density became Term Repetition Balance: repetition receives a penalty, never a reward.
Freshness logic was inverted
A stale visible date once outscored no date. Version 2 makes an absent date neutral and applies freshness only when the topic is time-sensitive.
Audit points looked like predicted gains
Projected-score language implied independent factor gains stacked. Controlled evidence showed they often do not, so 2.0 reports internal audit weights without forecasting citation lift.
Why 2.0 required a new baseline
The denominator changed, eleven factors became unscored diagnostics, new evidence-backed factors entered the score, and technical eligibility became a real gate. There is no honest formula for converting a 1.x score into a 2.x score because the measured model changed. The reason for a new baseline is valid comparison. Reusing the old score leads to a false trend.
Version 2 added four pipeline stages, evidence tiers on every factor, query-aware analysis, product and informational profiles, experimental engine presets, and citation-fitness gates. The tool still reads old configuration and JavaScript Object Notation (JSON) results. Continuous integration thresholds need fresh version 2 baselines.
Keep the old result as history, run a fresh 2.x baseline, and compare future changes only within a stable major version and query set.
Open the complete 2.0 migration guide ↗.
Key takeaways: the project today
aiseo-audit 2.0.1 uses the same analyzer for its command-line interface, TypeScript application programming interface, GitHub Action, and Model Context Protocol server. It produces terminal, JSON, Markdown, or self-contained HTML reports. Jeff Patterson and Agency Enterprise maintain the MIT-licensed project.
The current promise is limited. The tool makes observable AI search readiness easier to inspect and track. It cannot see a private engine index or know the sources in a future model answer.
- The prototype is the February 2026 geoaudit release.
- The package is the aiseo-audit project published on npm.
- The 1.x line is the period of rapid feature and scoring expansion.
- The 2.0 release is the evidence-tiered pipeline rebuild.
- The baseline is valid only within a stable major version and query set.
Bottom line: version 2 changed the scoring model enough to require a new baseline rather than a converted 1.x score.
Sources for the version history
According to the npm archive, the package records each published version. According to GitHub, the release notes document the feature timeline. According to the migration guide, version 2 requires new baselines. According to the evidence map, scored factors now record their evidence tier and pipeline stage.
- aiseo-audit version history: npm package registry.
- aiseo-audit releases and changelog: GitHub.
- Migration guide for aiseo-audit 2.0: project documentation.
- aiseo-audit evidence map: factor sources, tiers, and stages.