Preserve what was actually published
Capture the page title and meta description with the source URL and time. Keep raw observed values separate from a cleaned display version. If metadata is absent, a public profile can use a careful summary of visible content, but it should retain the origin of that summary. Never present an AI-generated sentence as the original site metadata.
Separate a site’s claim from an observed fact
A site may describe itself as the largest provider in its field. That is a marketing claim unless independently supported. An Atlas summary can say what the company offers without repeating unverified rankings, customer counts or certifications. The objective is a useful original description, not a copied advertisement or an authoritative endorsement of everything found in the page.
Write for the actual page
Useful descriptions summarize a distinct page in human language. Google may generate a different search snippet from page content rather than use the meta description verbatim. It does not publish one fixed character limit that guarantees a particular snippet. [E08] An analysis can flag missing or duplicated metadata without pretending to predict the exact search result a user will see.
Review the generated presentation
Compare the proposed summary with visible services and content. Remove unsupported claims and preserve names without translating them into something else. Keep owner-requested corrections distinct from automatically recrawled metadata. A later crawl should not silently overwrite a reviewed editorial correction. The public profile needs a history of what changed and why.
Use metadata as a source, not as an endorsement
A title and description tell us how a website presents itself. They may be outdated, duplicated, promotional or absent. Preserve the observed values with their source URL and timestamp; write any summary as a separate derived description. Do not turn a phrase such as “the most trusted provider” into an independently verified fact.
Check whether the title describes the actual page and whether important language variants have their own useful metadata. A technical profile can identify missing or inconsistent fields without claiming that the business behind the site is reliable, licensed or commercially successful.
| Decision | Useful requirement | Evidence |
|---|---|---|
| Observed title | Preserve the source value. | URL, extraction and date. |
| Summary | Separate interpretation from the source. | Linked evidence and cautious wording. |
| Recommendation | Describe a specific metadata issue. | Page-level example and proposed correction. |
Where this goes wrong
Never render extracted HTML or scripts directly inside a profile. External text is untrusted data. A missing description also does not justify inventing a company biography; the profile can stay factual using the visible content and explicit gaps.
A good profile explains the website without pretending to know more than the evidence supports.
Turn it into a working checklist
- Keep source URL and capture time.
- Separate raw metadata and summary.
- Avoid unsupported marketing claims.
- Preserve reviewed corrections.
A concrete next step
Keep observed metadata, derived summaries and owner-submitted corrections in distinct fields. This makes later review possible without erasing what the scanner actually saw.
Watch a related case
The video could not be loaded. Try again later.