Aleyda Solis, the SEO and AI search consultant behind Orainti, has published an updated free spreadsheet that turns Google’s public quality guidance into 83 actionable checks. The October 2026 edition spans 10 areas, and 76 of the checks point to 11 Google documents.
The update comes with a new Claude Skill that walks through the questions and writes an evidence-backed assessment. It is a working tool, not a shortcut to a Google score.
From guidance to operation
Each row is a question. Ten columns capture the area, the check, when it applies, how to verify it, the source text, the primary source, an assessment, evidence, a benchmark against the top result and review scope.
The assessment column has five values: Not assessed, Meets, Partly meets, Does not meet and Not applicable. There is no numeric score and no weighting. The practical engine is the evidence column, which asks for a URL, a record or a reason a check does not apply.
- Trust and authorship is the largest area with 13 checks; editorial validation has two.
- 54 checks apply to individual pages, 12 to the whole site, nine to page templates and eight to workflows.
What changed in Google’s own documents
The sheet captures documentation changes made in the five weeks before the update. On October 1, 2026, Google told sites to manually fact-check all AI-generated content before publishing and extended the requirement to titles, meta descriptions, structured data and image alt text. A day later, Google added main content, effort, talent or skill and accuracy sections to its people-first content page. Each change has a matching row in the checklist.
The update also reflects the August 2026 regional shift for site reputation abuse manual actions in the European Economic Area and Google’s May 2026 clarification that spam policies cover generative AI responses in Search.
The page experience section has the only numeric thresholds—Largest Contentful Paint of 2.5 seconds or less, Interaction to Next Paint at 200 milliseconds or less, and Cumulative Layout Shift at 0.1 or less, all measured at the 75th percentile of field data. It insists on real-user data; if field data is thin, mark the check “Not assessed.”
A Claude Skill, with its own limits
The skill is version 1.2.1, named eeat-helpfulness-assessment. It contains 27 files, including a checklist copy, a data file, presets and worked examples. Two examples are built around zero findings.
Its instructions are explicit: “Zero problems is a valid outcome.” Reports must state near the top that they are an expert assessment of the evidence, not a Google rating, a certification or a prediction of rankings or AI citations.
Why this matters for content and SEO teams
This checklist matters less for any single question than for how it changes the workflow. Because every check needs evidence, quality stops being a matter of taste and becomes a shared document across SEO, editorial and engineering.
The source labels also keep the tool honest. Seventy-one checks are Google Search guidance, five cite the rater guidelines, and seven are Solis’s own practical recommendations. That separation tells your team when a requirement is Google’s and when it is an editorial addition.
A useful starting point is the 13 trust and authorship checks. For high-stakes or YMYL pages, ask the team to attach original photos, records or observations behind first-hand claims, and to link author bios to relevant experience. That disciplines the review without pretending the result is a ranking signal.
Source: PPC Land
