Exceed quality threshold

Exceed Quality Threshold

Purpose

Determine what a page must do to become eligible and preferable for a connected query network. Compare like with like, identify the demonstrated lower and upper boundaries, and require a material usefulness difference rather than superficial imitation.

Treat "quality threshold" as a comparative SEO model, not a disclosed Google score or ranking factor. Separate observations from hypotheses and recommendations. Read references/evidence-and-guardrails.md before making claims about Google, Koray's framework, or prescriptive tactics.

Required inputs

Collect or infer:

• Primary query or topic
• Target page, draft, brief, or proposed resource
• Publishing site and intended source type
• Target market, language, device, and current date
• Intended audience and conversion/task goal
• Any first-party evidence, expertise, data, or tools available

Ask only for a missing input that would materially change the cohort or verdict. If no page exists, evaluate the proposed format and produce a build specification. If the site is unavailable, continue with a provisional page-level assessment and label source-level findings unknown.

Evidence rules

1. Use live web search whenever available. SERPs, freshness, page contents, and reputations change.
2. Record query, market, language, device when known, and date. Do not imply a neutral or universal SERP.
3. Prefer direct inspection of ranking pages over snippets. Snippets cannot establish completeness, evidence, or page experience.
4. Cite the pages used for decisive observations. Distinguish observed facts, reasonable inference, and unknowns.
5. Never invent rankings, page contents, traffic, links, authority, authorship, methodology, or first-party evidence.
6. If live search or page access is unavailable, use user-supplied exports or URLs and label the result provisional; do not claim to have measured the current threshold.
7. Do not equate length, heading count, entity count, schema, images, or n-gram overlap with quality. Count an element only when it helps satisfy the task or establishes evidence.

Workflow

1. Define the task and query network

State the searcher's core job in one sentence. Build approximately 8–15 closely connected queries, adapting the number to topic breadth:

• Main query and close reformulations
• Definition or explanatory variants
• How, why, which, comparison, and decision variants
• People Also Ask and related-search questions
• Logical searches immediately before and after the main task

Exclude variants with materially different audiences or tasks. Mark likely freshness-sensitive queries.

2. Map the live result set

Inspect up to the first 20 organic results for the main query and enough results across the network to find recurrence. Do not count ads or SERP features as organic positions. Record for each useful result:

Field	Record
Query and rank	Where the result appeared
Search intent	Learn, compare, buy, calculate, find data, or complete a task
Page type	Report, guide, category, product, tool, forum, or video
Source type	Specialist, company, marketplace, government, general media, community, or other
Primary answer	What the result actually gives the searcher
Evidence	Data, citations, experience, methodology, author qualification
Format	Tables, charts, steps, calculator, video, download, discussion
Freshness	Publication/update date and whether recency matters
Recurrence	Number and breadth of connected queries where the source appears
Special support	Exceptional brand demand, links, reputation, or navigational intent when evidenced

When result collection is incomplete, report the coverage achieved rather than silently reducing the sample.

3. Select the comparator cohort

Choose the cohort matching the intended page and source identity. Compare a research report primarily with research publishers, a product category with comparable commercial pages, and a tool with tools that complete the same job.

Use intent and task completion as the first filter, then page type and source type. Do not combine publishers, ecommerce stores, forums, videos, and research reports into one undifferentiated benchmark merely because they share a SERP.

Explain:

• Included cohort and why it is comparable
• Excluded result types and why
• Any mixed-intent ambiguity

If fewer than three credible comparators exist, widen only to the nearest task-equivalent cohort and lower confidence.

4. Find the quality range

Within the chosen cohort identify:

• Lower boundary: weakest credible result that recurs or maintains meaningful first-page visibility
• Upper boundary: strongest representative result across several connected queries
• Outliers: results whose visibility appears unusually dependent on exceptional brand, authority, links, freshness, or navigational demand

Recurrence is evidence of representativeness, not proof of causation. Do not choose boundaries from a single query when network evidence is available. Exclude outliers from the content benchmark but document them.

5. Build the information-unit matrix

Extract useful, verifiable answer components rather than keywords or shared phrasing. For every unit, record whether each comparator provides it, the quality of that treatment, and supporting URL.

Classify each unit:

• Entry requirement: common in qualifying results and necessary to complete the task
• Best-in-class characteristic: concentrated in strong representatives and materially useful
• Information gap: requested or logically needed but poorly answered across the cohort
• Imitation noise: common but not demonstrably useful

Include definitions, scope, dates, methodology, limitations, examples, decision criteria, and retrieval format when relevant. Do not infer necessity solely from frequency.

6. Audit the target through five gates

Rate every gate pass, conditional, fail, or unknown. Give evidence and a specific remedy; do not collapse the result into a pseudo-precise score.

Intent and format

• Perform the same core task as the cohort
• Surface the primary answer immediately
• Use the appropriate page format
• Resolve the dominant interpretation without mixing unrelated contexts

Information

• Cover every necessary information unit accurately
• Use specific entity attributes rather than vague statements
• Follow the searcher's likely sequence of questions
• Fill at least one valuable gap

Evidence and originality

• Support consequential claims with appropriate primary evidence
• Make definitions, dates, sample sizes, and limitations explicit
• Make methodology reproducible where applicable
• Contribute original data, experience, analysis, synthesis, or tooling

Originality must be substantive. Novel wording for commodity information does not qualify.

Source qualification

• Demonstrate site focus and topical consistency
• Name suitably experienced authors or reviewers
• Connect relevant supporting pages coherently
• Keep declarations and definitions consistent across pages
• Make the organization, ownership, purpose, and reputation legible

Mark site-level items unknown when they cannot be verified. A page can be excellent while its source qualification remains a constraint.

Experience and retrieval

• Make the answer fast to scan and understand
• Prefer useful tables, charts, steps, or tools over equivalent prose
• Support mobile use and accessibility
• Separate intrusive promotions from the main content
• Verify indexability, canonical handling, and sensible internal links when technical access exists

7. Require a relative quality difference

Define the smallest build that clears all entry requirements, then define why a searcher or citation source would prefer it over the upper boundary. Require at least one defensible advantage that the existing cohort cannot cheaply replicate, such as:

• First-party or newly assembled data
• Reproducible methodology and inspectable evidence
• Demonstrable hands-on experience
• Better task completion through a tool, template, calculator, or downloadable dataset
• A meaningful segment, time-series, or comparative analysis competitors omit
• Faster and clearer retrieval without loss of completeness

Reject "longer," "more headings," "more entities," "add FAQ schema," or "rewrite competitors" as standalone differentiators.

8. Issue the verdict and action plan

Use one verdict:

• Exceeds: clears all material gates and has a defensible advantage over the upper boundary
• At threshold: meets entry requirements but lacks enough relative difference to deserve preference
• Below threshold: misses one or more entry requirements
• Indeterminate: evidence or comparator coverage is too weak for a responsible decision

Prioritize actions as:

1. Eligibility blockers
2. Evidence and trust deficits
3. Original opportunity
4. Retrieval and presentation improvements
5. Optional enhancements

Make each action testable. Name the information unit, evidence, format, owner/input needed, and acceptance condition where practical.

Output format

# Quality-threshold assessment

**Verdict:** [Exceeds / At threshold / Below threshold / Indeterminate]
**Confidence:** [High / Medium / Low]
**SERP context:** [market, language, device if known, date]
**Core task:** [one sentence]

## Comparator cohort
[Included types, excluded types, and rationale]

## Quality range
| Boundary | Representative | Recurrence | Why it qualifies |
|---|---|---:|---|
| Lower | ... | ... | ... |
| Upper | ... | ... | ... |

## Information-unit matrix
| Unit | Lower | Upper | Target | Classification | Evidence |
|---|---|---|---|---|---|

## Five-gate audit
| Gate | Status | Evidence | Required change |
|---|---|---|---|

## Relative quality difference
[Why the target deserves preference, or why it currently does not]

## Prioritized action plan
1. ...

## Limits
[Coverage gaps, inaccessible pages, causal uncertainty, and source-level unknowns]

Keep the matrix proportional to the decision. Include the decisive units and provide an appendix when the full query/result inventory would overwhelm the recommendation.

Failure modes

• Do not claim Google uses one measurable score named "quality threshold."
• Do not treat correlation in ranking pages as causation.
• Do not benchmark the target against incompatible result types.
• Do not select the weakest result as the quality goal; use it only as an eligibility boundary.
• Do not copy headings, wording, or content patterns without explaining their user value.
• Do not recommend fake authorship, fabricated experience, unsupported claims, or decorative citations.
• Do not prescribe EXIF preservation, arbitrary refresh percentages, fewer links, Q&A structure, FAQ schema, or entity/n-gram maximization as universal rules.
• Do not promise rankings. Frame the output as a comparative content and source-quality decision.

Companion skills

• Use authority-mark-keyword-difficulty separately when the user needs a domain-authority access signal. Authority feasibility does not replace this content-quality analysis.
• Pass a below-threshold new-page plan to seo-content-brief after defining its mandatory units and original opportunity.
• Pass an existing underperforming page to content-refresh-brief when the required action is a refresh specification.