The Nine Signals
Freshness
Whether assistants see current facts and current evidence when they answer local business prompts.
What it measures
Freshness is not one timestamp. It has four layers:
| Layer | What it means |
|---|---|
| Entity freshness | The business fact itself is current, such as hours, phone, service availability, or temporary closure. |
| Evidence freshness | The cited page, review, profile, or article is recent enough for the claim. |
| Index freshness | The updated fact has reached the search or assistant corpus. |
| Answer freshness | The assistant's final answer reflects the newest valid evidence. |
Why it moves your EAC
Freshness matters most when the user's question has a clock: open now, holiday hours, emergency service, same-day appointment, current menu, current price, active offer, or recent reviews.
A business can fix its website and still fail the answer if the index, listing provider, or assistant answer layer has not caught up.
What the agent sees
Weak read
The website has updated holiday hours, but Apple, Bing, or a directory still lists old hours, and the assistant cites the stale source.
Strong read
Website, schema, profiles, sitemap signals, and major listings all reflect the new hours, and assistants answer correctly after propagation.
How to improve
- Maintain one source of truth. Track hours, special hours, closure status, services, booking URLs, menus, and contact changes centrally.
- Push changes to platforms. Update Google Business Profile, Apple Business, Bing Places, major directories, and the website together.
- Use meaningful update signals. Update structured data, sitemap
lastmod, and IndexNow where supported after real content changes. - Keep reviews recent and authentic. Recent reviews provide time-aware trust signals, especially in competitive local categories.
- Retest by fact volatility. Daily for volatile facts, weekly for high-risk categories, monthly for full provider comparison, and before major holidays.
How to verify
How long did the answer take to correct?
Track change-to-answer lag, stale-answer rate, stale-citation share, and provider-specific bottlenecks after every material update.
Re-run checkResearch basis: docs/research/deep-research-report(13).md