AI Editorial Policy

Last reviewed 2026-07-22

ONULSURI publishes independent AI guides for readers who need practical selection help. This policy explains how editorial content is reviewed, what verification means, and how corrections are handled.

The AI section is separate from calculators and CMS drafts. Editorial pages are static, source-aware where relevant, and intentionally avoid fabricated prices, rankings, or benchmark scores.

Review process

Each published AI page has a last-reviewed date and a next-review due date derived from a fixed review cadence.

Editors check whether claims still match the product surfaces they describe, whether related links resolve, and whether notices about pricing or benchmarks remain accurate.

Confirm the page still describes the intended product or topic scope.
Re-check related Compare, Tool, Pricing, Benchmark, Guide, and Prompt links.
Update verification status when evidence is incomplete or outdated.

Update cadence

Cadence varies by section. Pricing pages move faster because vendor plans change often. Guides, benchmarks, and prompts usually follow a longer interval because they describe frameworks rather than daily product diffs.

Cadence is deterministic: next review due is computed from last reviewed plus a fixed interval. Production pages do not invent “today” from wall-clock time.

Pricing: monthly by default.
Compare and Tools: quarterly by default.
Guides, Benchmarks, and Prompts: every six months by default.
Methodology pages may use a longer interval when the framework is stable.

Source standards

Where product facts matter — especially pricing — editors prefer official vendor documentation over secondary summaries.

Comparisons and guides may synthesize public product behavior, but they should not invent capabilities, prices, or evaluation scores.

Verification levels

Verification status communicates how thoroughly a page was checked against available evidence at last review.

Verified: core claims checked against expected sources or editorial criteria for that section.
Partially verified: some claims checked; remaining items noted as caveats or pending review.
Needs review: the page should not be treated as freshly confirmed.
Stale: the page is known to be overdue for a substantive update.

Corrections policy

Material errors — wrong product attribution, broken selection guidance, or outdated plan structure that could mislead a purchase decision — should be corrected in the next editorial pass and reflected in the last-reviewed date.

Minor wording improvements do not always require a cadence reset, but factual corrections do.

Benchmark methodology references

Benchmark pages describe evaluation frameworks and scenario suites. They do not publish product winners or fabricated score reports.

For the shared outcome rubric and evidence expectations, start with the AI Benchmark Methodology page.

Pricing methodology references

Pricing pages avoid fabricated dollar amounts and point readers to official vendor sources.

For source rules, review freshness expectations, and what we refuse to invent, read the AI Pricing review methodology.

FAQ

Does this policy apply to every AI page?

Yes. Compare, Tool Detail, Pricing, Guides, Benchmarks, and the Prompt Library share the same editorial governance layer, with section-specific cadence and evidence defaults.

Why do pages show published and last-reviewed dates?

Published marks when the page entered the library. Last reviewed marks the most recent editorial check. Next review due is derived from the section cadence.

Are verification badges product endorsements?

No. They describe editorial confidence in the page content, not a recommendation to buy or adopt a product.

Where can I read pricing and benchmark rules?

Use the pricing methodology for plan-structure rules and the benchmark methodology for evaluation frameworks. This policy explains how those pages fit the broader editorial process.