Content freshness signals for AI search: how to date-stamp content AI engines trust

By Stephan Charles | Last fact-checked: <time datetime="2026-08-06">2026-08-06</time>

Half of all AI citations go to content published or updated within the past 13 weeks, according to Lily Ray's Amsive research analyzing AI-cited URLs across ChatGPT, Perplexity, and Google AI Overviews. Content updated within 30 days earns 3.2x more citations than content older than 90 days. Five technical signals determine whether AI engines classify your page as fresh or stale: the dateModified field in Article schema, the lastmod tag in your sitemap, the visible date on-page, the HTTP Last-Modified header, and in-body date references. Each one is fixable without a developer.

BrandCited is an AI brand visibility platform that tracks citations across 8 engines and scores each brand on a 0-100 composite index. Freshness defects — missing or outdated dateModified fields, sitemap lastmod tags that do not match actual updates, on-page dates from 2023 — appear in BrandCited's audits as P1 defects across 67% of tracked brands. Run a free freshness audit at brandcited.ai to see which of your pages carry stale signals.

<section id="why-does-ai-search-favor-fresh-content">

Why does AI search favor fresh content over highly ranked old content?#

AI search engines weight recency because answer accuracy degrades as information ages. A Perplexity answer citing a 2023 pricing page is citing outdated information — and Perplexity knows it. Retrieval-augmented systems penalize stale sources to protect answer quality, not to reward recency for its own sake.

Ahrefs analyzed 17 million AI citations and found that AI-cited content is 25.7% fresher on average than content cited in traditional organic search results. That gap holds across query types. For time-sensitive queries — pricing, comparisons, current events — the recency window narrows further: ChatGPT prefers content from the last 30 days for those queries, not the last 90.

Content has a 1-year half-life in AI search. Gander's freshness analysis found that a page loses roughly 50% of its AI citation potential within 12 months of publication — even if its Google organic ranking stays stable. The mechanisms differ: Google's ranking algorithm weights content authority accumulated over time, while AI retrieval systems weight whether this source reflects the current state of the topic.

The practical consequence: 65% of AI bot crawl hits target content published within the past year, and 89% target content updated within the past three years. Pages published before 2024 and never updated are competing from a structural disadvantage in every retrieval-augmented system.

</section>

<section id="what-are-the-5-freshness-signals">

What are the 5 freshness signals AI engines actually read?#

AI engines read freshness from five distinct technical signals per page. Missing one does not kill citations, but the strongest citation lift comes from all five matching.

Signal 1: dateModified in Article JSON-LD. The dateModified property in your Article schema block is the primary machine-readable freshness signal. AI crawlers parse JSON-LD before they parse prose. A page with "dateModified": "2026-08-06" in its schema is explicitly telling every crawler when the content was last changed. Per Stackmatix's structured data guide, this field is weighted by Perplexity, Bing Copilot, and Google's AI Overview system when they pick between two pages covering the same topic.

Signal 2: Sitemap lastmod tag. The <lastmod> element in your sitemap.xml tells crawlers when to reprioritize a page for recrawl. A 30-day server log study by Digital Applied found that ClaudeBot's standard revisit cycle is 6.8 days, but when a fresh Last-Modified header is detected, that cycle drops to 1.6 days. Updated sitemap lastmod dates after a content change cause ClaudeBot and PerplexityBot to recrawl within two weeks instead of seven.

Signal 3: Visible on-page date. AI crawlers parse rendered HTML, not just schema. A visible <time datetime="2026-08-06">Updated August 2026</time> element in the article header confirms to the crawler that the dateModified claim matches what a human reader would see. When the schema date and the visible date conflict, retrieval systems flag the mismatch as a trust signal failure.

Signal 4: HTTP Last-Modified header. Your server sends an HTTP Last-Modified header with every page response. Most CMS platforms update this automatically when a page is re-saved. For static sites, this requires a deployment pipeline that touches the page's file modification time on publish. Crawlers use this header as a pre-parse signal before downloading and processing page content.

Signal 5: In-body date references. AI engines parse prose for date signals as well as structured metadata. A sentence like "as of Q3 2026, Perplexity processes 1.2 billion queries per month" carries a freshness signal in the prose itself. Conbersa's freshness research found that embedding current-year statistics and references to recent events within article bodies drives stronger citation lift than updating metadata alone — because the body text is what retrieval systems quote.

</section>

<section id="what-counts-as-meaningful-update">

What counts as a meaningful content update that AI engines recognize?#

A meaningful update changes the substance of the content, not just its date. This distinction matters because AI engines and Google have both flagged date inflation — changing the dateModified field without updating content — as a trust issue. A page that claims to be updated today but contains only 2023 information fails the trust check retrieval systems run.

Three types of updates qualify as meaningful:

Replace outdated statistics with current ones. If your article says "ChatGPT has 400 million users" and the current figure is 800 million, swap it. Link the new figure to a dated source. The in-body date reference, the updated statistic, and the dateModified field all align — and the alignment is what AI systems verify. A single stat swap with a new source link satisfies the freshness check for a quarterly update cycle.

Add a new section that answers a question the original missed. Foglift's content freshness checklist identifies this as the highest-impact update type for AI citations. AI engines that retrieve content based on query matching get more retrieval surface from an article that answers six questions than one that answers four. Adding a section addressing a new sub-question that the topic has generated since publication is the most efficient way to increase both freshness and coverage.

Correct stale claims and broken links. A claim that was accurate in 2024 and is now wrong is an active citation liability. When AI engines encounter outdated claims in retrieved content, they downweight the source. An annual pass through your top-ten articles to correct stale claims is maintenance that pays in sustained citation rates.

What does not count: changing the date field without touching content, reformatting bullet points, or updating the page title without altering any prose. These changes may update HTTP Last-Modified headers but do not change the content that AI systems extract and verify.

</section>

<section id="datmodified-json-ld-template">

How do you implement dateModified correctly in Article JSON-LD?#

The dateModified field goes in the Article schema block in your page's JSON-LD. Here is the minimum-viable Article schema block with correct freshness fields:

Four rules for this block:

1. 1datePublished stays fixed at the original publication date. Changing it after the fact causes trust issues in Google Search Console.

1. 1dateModified updates every time you make a substantive content change. It uses ISO 8601 format: YYYY-MM-DD. It must match the <lastmod> in your sitemap.xml and the visible date on the page.

1. 1The author block must include sameAs with a link to a verifiable profile. Perplexity weights author byline prominence as a citation signal.

1. 1The JSON-LD block goes in the <head> of the page. Placing it in the body does not break parsing but is not the canonical placement AI crawlers expect.

Pair this with an updated sitemap entry:

The lastmod value in sitemap.xml should match dateModified in your JSON-LD. Mismatches between these two values are a documented signal of date manipulation that reduces retrieval system trust.

</section>

<section id="which-engines-most-sensitive-to-freshness">

Which AI engines are most sensitive to freshness, and how often should you update?#

The 8 major AI engines weigh freshness differently. Update cadence should match the engine with the shortest recency window, not the average.

Perplexity is the most freshness-sensitive engine. Approximately 50% of all Perplexity citations come from current-year content alone, per Amsive's research. Perplexity runs live web retrieval on every query, so a page updated yesterday competes with a page last updated two years ago on equal technical footing — but the 2-years-old page loses the freshness tie-break. For brands that rely on Perplexity citations, monthly updates to their top-five pages are the minimum viable cadence.

ChatGPT in web-browsing mode shows the most aggressive 30-day recency bias: 76.4% of its most-cited pages were updated within the past 30 days. ChatGPT's standard (non-browsing) mode draws on parametric memory with a training cutoff, so freshness signals only affect ChatGPT citations when it is browsing the web for the query. For time-sensitive queries — pricing, comparisons, product specs — browsing mode activates by default, and fresh pages win.

Google AI Overviews has the weakest freshness bias of the three, with citation patterns that track more closely with traditional organic ranking age profiles. Ahrefs' 17M citation study found AI Overviews cite content 25.7% fresher than organic results on average — but Google's Knowledge Graph authority still outweighs recency for established brands.

ClaudeBot (Anthropic's search indexing crawler) revisits pages that show fresh Last-Modified headers on a 1.6-day cycle, per the Digital Applied log study. Fresh signals from ClaudeBot lead to faster inclusion in Claude's retrieval pool for web-augmented responses.

Update frequency guide by content type:

</section>

<section id="brandcited-freshness-audit">

How does BrandCited audit freshness signals?#

BrandCited's audit engine checks five freshness signals per page as part of its 30+ check scan: dateModified in Article schema, sitemap lastmod tag, visible on-page date, HTTP Last-Modified header, and in-body date references. Pages where these signals are absent, mismatched, or older than 90 days are flagged as freshness defects — P1 severity, meaning they suppress citation rates without blocking them entirely.

Freshness defects appear in 67% of BrandCited's tracked brands. The most common pattern: a blog post published in 2024 with no dateModified field, no sitemap lastmod update after a minor edit, and a visible date header that says only the original year. From every AI crawler's perspective, that page is 18 months old regardless of when the last edit was made.

Run a free BrandCited scan at brandcited.ai to see which of your pages carry freshness defects and get a prioritized fix list ranked by citation impact.

</section>

What to do right now#

1. 1Add dateModified to every Article schema block today. Check your top 10 blog posts and landing pages. If any are missing dateModified in their JSON-LD, add it. If the value is older than 90 days, update the content first, then update the date. Use the template above — it takes under 10 minutes per page.

1. 1Update your sitemap.xml lastmod after every content change. Most CMS platforms (WordPress, Webflow, Framer) handle this automatically if you configure your sitemap plugin correctly. For Next.js or custom static sites, add a sitemap generation step to your deployment pipeline that updates lastmod to the current date on changed files.

1. 1Put a visible updated date on every article. Place it near the author byline in the article header using semantic HTML: <time datetime="YYYY-MM-DD">Updated [Month Year]</time>. The on-page date confirms to AI crawlers that the schema date is accurate. This single change takes under 5 minutes per page and applies permanently in your template.

1. 1Run a quarterly content refresh cycle for your top 5 pages. Set a recurring quarterly task to update at least one statistic, fix one broken external link, and add one new section to your top-performing pages. Pages that go 6 months without updates start losing citation rates on Perplexity within that window.

1. 1Replace the oldest statistics in your most-cited pages first. Sort your pages by last-modified date. Any page with a statistic older than one year is a freshness liability. Replace that statistic with a current source link and update dateModified. This is the highest-impact single action for recovering citations on Perplexity, where 50% of citations come from current-year content.

1. 1Run a BrandCited freshness scan to find all defects at once. Rather than auditing 50 pages by hand, run a BrandCited scan at brandcited.ai to surface every freshness defect across your site in 30 seconds. The scan shows which pages are generating freshness penalties and ranks them by citation impact so you fix the highest-value pages first.

<section id="news-roundup">

AI search updates from the last 24 hours#

• OpenAI names Astra its next major model family: On August 1, 2026, OpenAI announced Astra — its next model family after GPT-5.6 — by publishing a 249-page mathematics manuscript with machine-checkable Lean 4 proofs for 10 previously unsolved problems. Astra is designed for multi-agent long-horizon tasks. No public release date yet; a 30-day federal pre-release review makes September the earliest plausible window. (The Decoder)

• GPT-5.6 Luna price drops 80%, Terra drops 20%: Effective July 30, 2026, OpenAI cut GPT-5.6 Luna pricing from $1.00 to $0.20 per 1M input tokens and Terra from $2.50 to $2.00. Luna's output price dropped from $6.00 to $1.20. These are permanent list price changes. (OpenAI)

• Perplexity Agent API goes GA: Perplexity's Agent API reached general availability in August 2026, now supporting Claude Opus 4.7, GPT-5.5, and Grok 4.20 Reasoning as callable models. This expands the number of developer pipelines that can route queries through Perplexity, increasing its citation surface. (Perplexity changelog)

• ChatGPT Business free tier ends today: ChatGPT Business usage that was free through August 6, 2026 now draws from the workspace credit pool under flexible pricing. Brands using ChatGPT Business for internal AI search workflows should audit their credit budgets today. (OpenAI release notes)

</section>

FAQ#

How often should I update content to stay visible in AI search?

ChatGPT cites pages updated within the last 30 days at 76.4% frequency. Perplexity pulls roughly 50% of its citations from current-year content. The practical update schedule is: product and pricing pages monthly, data-heavy comparison guides quarterly, and evergreen how-to content every six months. Each update must add new statistics, revised claims, or a substantive new section — changing only the date triggers a trust penalty, not a citation lift.

Does the dateModified field in JSON-LD actually affect AI citations?

The dateModified field in Article schema is the primary machine-readable freshness signal AI crawlers parse. A page with a visible on-page date of April 2026 and a matching dateModified field in its schema is chosen over an identical page with a 2023 date when retrieval-augmented engines pick between sources. The field must match the actual content update — inflating it without making substantive changes is a documented trust issue that search engines flag.

What is the 13-week rule for AI search citations?

Lily Ray, VP of SEO and AI Search at Amsive, analyzed which URLs get surfaced in AI-generated answers and found that 50% of all AI citations come from content less than 13 weeks old. Content under 30 days earns 3.2x more citations than content older than 90 days. Content older than one quarter is competing from a structural disadvantage in retrieval-augmented systems like Perplexity and Google AI Overviews.

Does ClaudeBot crawl fresher pages more frequently?

A 30-day server log study by Digital Applied found that ClaudeBot's standard revisit cycle is 6.8 days, but when a fresh last-modified header is detected on a page, ClaudeBot drops its revisit cycle to 1.6 days. Updating your sitemap.xml lastmod dates after each content update signals freshness to ClaudeBot and causes it to recrawl within two weeks instead of seven.

What counts as a meaningful content update for AI search?

A meaningful update replaces at least one statistic with a newer source, adds a new section that answers a question the original post missed, or corrects stale claims that have become inaccurate. Changing only the dateModified timestamp without altering content does not count and is flagged as a trust issue by Google. The freshness signal is strongest when dateModified, sitemap lastmod, on-page visible date, and HTTP Last-Modified header all match and all reflect a real content change.

Run a free AI visibility audit on your brand at brandcited.ai. You will see your freshness score across 9 AI platforms in 30 seconds, with every defect ranked by citation impact.