Table of contents
- AI Always Prioritizes the Most Current Sources
- Freshness in AI Works Differently Than in Traditional Google
- Comparison of Freshness Signals
- Recommended Update Frequency by Industry and Content Type
- Content Audit Workflow: 4 Steps to Find Articles That Need Updating
- Decision Matrix: Update an Old Article or Create a New One?
- JSON-LD Article with dateModified Used Correctly
- Case Study: Updating an Old Article Increases AI Citations
- Freshness Update Checklist Before Publishing
AI Always Prioritizes the Most Current Sources
When a user asks AI "What is the current home loan interest rate?", AI will not cite an article from 2022 — that data could be completely wrong relative to today's reality. AI is designed to recognize this and prioritize more recently updated sources.
But freshness isn't only important in finance or law. Even a "what is SEO" article written in 2021 gradually loses credibility once AI understands that the search landscape has changed significantly with the emergence of AI search. Guides covering schema markup, content strategy, or digital marketing are particularly susceptible to becoming outdated.
Content freshness is the most overlooked factor in AEO — yet it is one of the most common reasons articles gradually lose AI visibility over time, even when the underlying content quality remains high.
Freshness in AI Works Differently Than in Traditional Google
With traditional Google Search, freshness primarily works through the "Query Deserves Freshness" mechanism — a small category of queries for which Google prioritizes recent results (breaking news, events, newly launched products). Most queries are not significantly affected by freshness.
With AI, the mechanism is different and broader: AI synthesizes information from multiple sources simultaneously and independently evaluates the trustworthiness of each based on multiple signals — freshness being one of them. If source A and source B both correctly answer a question, but A was updated 2 months ago and B was updated 18 months ago, AI will typically favor A — especially for topics that are likely to change over time.
An important point to remember: newer is not automatically better. Content that is fresh but inaccurate will be dismissed by AI. Freshness only works in your favor when it is accompanied by quality and factual accuracy.
Comparison of Freshness Signals
| Signal | How AI/Search Reads It | Level of Impact | Can It Be Faked? |
|---|---|---|---|
dateModified in Article schema | Read directly from JSON-LD | Very high | Yes — but detected if content didn't actually change |
Last-Modified HTTP header | Server response header | High | Yes — but usually automatic based on the file system |
lastmod in sitemap.xml | Crawler reads during crawl | Medium | Yes — but has little effect if content didn't change |
| Actual content change | Crawler compares against previous crawl | Very high | No — this is the most objective and trustworthy signal |
datePublished | Read only, cannot be changed | Low for freshness | Should not be — this is the original publication date |
Recommended Update Frequency by Industry and Content Type
| Industry/Content Type | Recommended Frequency | Reason |
|---|---|---|
| Finance, banking | Monthly | Interest rates and policies change frequently |
| Legal | With each new decree or regulation | Citing outdated law completely destroys credibility |
| Technology, digital marketing | Quarterly | New tools and algorithm updates are constant |
| Health, medical | Quarterly | Treatment guidelines and recommendations can change |
| Interior design, construction | Every 6 months to a year | Trends and materials change more slowly |
| Culture, history | 1–2 years | Rarely changes, update only when new research emerges |
| Technical how-to guides | Quarterly | Tools and platforms change; screenshots go stale quickly |
| "What is" articles (evergreen) | Annually | Add new FAQs, update examples and data |
Content Audit Workflow: 4 Steps to Find Articles That Need Updating
Step 1 — Export data from Google Search Console: Go to Performance → filter by the most recent 6-month period vs. the 6 months before that. Export a list of articles with: impressions down more than 30%, CTR declining while impressions remain stable (a sign the content is no longer a good match), or rankings that dropped from the top 5 to the top 10.
Step 2 — Filter by priority criteria: Flag articles that meet at least 2 of 4 conditions: last updated more than 12 months ago; contains specific data or legislation; falls under YMYL categories; was previously cited by AI (you can check by asking AI directly) but citations have recently declined.
Step 3 — Classify the type of update needed: For each article in the list, identify what kind of update is required: data update only, adding a new section, updating laws/policy, adding new FAQs, or updating examples and case studies.
Step 4 — Prioritize and schedule: Build a table with columns: URL, update type, estimated effort (hours), expected impact (high/medium/low), deadline. Prioritize high-impact, low-effort articles first — these are typically how-to guides that just need a screenshot update or new data points.
Decision Matrix: Update an Old Article or Create a New One?
| Situation | Decision | Reason |
|---|---|---|
| Same topic, same intent, outdated content | Update the old article | Preserve the URL's accumulated authority and backlinks |
| Similar topic but different audience | Create a new article | Different intent requires a separate page |
| Old article is too short and lacks depth | Update and substantially expand | Freshen and improve quality at the same time |
| Old article covers a completely obsolete topic | 301 redirect to new article | Don't retain content that no longer has value |
| Old article ranking in top 5 but outdated | Update carefully | Don't risk disrupting an existing ranking |
| A sub-topic not yet covered on the site | Create a new article | New topic needs its own URL and authority |
JSON-LD Article with dateModified Used Correctly
{
"@context": "https://schema.org",
"@type": "Article",
"@id": "https://aeosaigon.com/blog/faq-schema-la-gi",
"headline": "FAQ Schema là gì? Hướng dẫn khai báo để xuất hiện trên Google và AI",
"description": "Hướng dẫn toàn diện về FAQPage schema: cách hoạt động, cách khai báo đúng chuẩn và lý do tại sao AI ưu tiên trích dẫn trang có FAQ schema.",
"datePublished": "2025-03-15T08:00:00+07:00",
"dateModified": "2026-07-20T10:30:00+07:00",
"author": {
"@type": "Person",
"name": "Nguyễn Thành Phát",
"url": "https://aeosaigon.com/tac-gia/nguyen-thanh-phat"
},
"publisher": {
"@type": "Organization",
"@id": "https://aeosaigon.com/#organization",
"name": "AEO Saigon",
"logo": {
"@type": "ImageObject",
"url": "https://aeosaigon.com/images/logo.png",
"width": 200,
"height": 60
}
},
"image": {
"@type": "ImageObject",
"url": "https://aeosaigon.com/images/faq-schema-la-gi.jpg",
"width": 1200,
"height": 630
},
"mainEntityOfPage": {
"@type": "WebPage",
"@id": "https://aeosaigon.com/blog/faq-schema-la-gi"
},
"articleSection": "Hướng dẫn",
"inLanguage": "vi",
"keywords": ["FAQ schema", "schema markup", "rich results", "AEO"]
}
Important note: datePublished is the date the article was first published and must never be changed. dateModified is the date of the most recent update — this is the field you need to update every time you make substantive changes to the content.
Case Study: Updating an Old Article Increases AI Citations
A guide on "Schema markup for blogs" was written in early 2025 and began losing visibility by year-end — AI citations fell 40% and Google impressions dropped 25% over three months.
After an audit, the content team identified 4 specific issues: the JSON-LD examples in the article used outdated syntax that was no longer recommended; there was no mention of newly prominent schema types (FAQPage, HowTo); there was no section on Next.js App Router implementation; and the rich results CTR data was sourced from a 2023 study.
After three weeks of comprehensive updates — adding 2 new sections, correcting all JSON-LD examples, updating data from the latest sources — dateModified was changed to the actual date the updates were completed. No URL change; no new article created.
Results over the following 6 weeks: Google impressions up 45%, CTR up 20%, and AI citations doubled compared to before the update. All backlinks from the original article were preserved.
Freshness Update Checklist Before Publishing
- At least 3 substantive changes in the content — not just typo fixes
- Update
dateModifiedin Article schema to the actual update date - Keep
datePublishedunchanged — never modify it - Add a note such as "Updated [Month/Year]" at the top or bottom of the article
- Update
lastmodinsitemap.xml - Purge the CDN cache after deploying
- Request Indexing in Google Search Console
- Record in the audit spreadsheet: update date and types of changes made
- Set a reminder to review the article again in 6–12 months depending on the industry
- Monitor impressions and CTR in Search Console 4–6 weeks after the update to measure impact
Frequently asked questions
How does AI assess content freshness?
3 primary signals: (1) dateModified in Article schema — the most important, AI reads it directly from JSON-LD; (2) Last-Modified HTTP header — the server returns the date the HTML file was last modified; (3) Content signals — AI detects real changes in the content such as new data, new sections, or updated laws and regulations. Note: you should not fake dateModified without making real changes to the content — AI is getting increasingly good at detecting 'touched but not updated'.
How old does content need to be before AI stops citing it?
There is no fixed threshold — it depends on the industry and type of query. Fast-changing industries (finance, law, tech): AI strongly favors content under 6 months old. Stable industries (technical guides, interior design): content 1–2 years old can still be cited if it remains accurate. Queries asking for 'the latest' or 'current' information prioritize freshness strongly; queries asking 'what is' or 'how to' prioritize accuracy over freshness.
What updates to an article make AI recognize freshness?
5 types of updates with genuine value: (1) New data — replace old statistics with the latest data and cite the source; (2) New sections — add an H2/H3 covering an aspect not previously addressed; (3) Updated laws or regulations; (4) New FAQs — add questions that customers have actually asked in the past 6 months; (5) New case studies. Fixing typos or changing CSS alone does not constitute meaningful content freshness.
Should I create a new article or update an old one?
Rule of thumb: if the search intent is the same, update the old article (keep the URL, accumulate authority). If the intent is different, create a new article. Example: 'FAQ schema 2026' — update the existing 'what is FAQ schema' article. 'FAQ schema for WooCommerce' — create a new article because the intent and audience are different. Old articles have backlinks and authority — updating is more effective than creating new in 90% of cases.
AEO Saigon
An Answer Engine Optimization agency in Ho Chi Minh City — helping business websites get cited by AI. About AEO Saigon →
Want your website to be cited by AI like this?
Free Audit