AI Visibility Watch #6 — July 17, 2026: The Citation-Freshness Penalty Is Real
Founder & CEO, Crackle PR
Something quiet is happening in the citation graph
Across the client prompt baskets we track, pages with a dateModified older than roughly 18 months are getting cited noticeably less than they were in Q1. The pages are still indexed. They still rank in classic SERPs. But when ChatGPT Search or a Google AI Overview picks a source to quote, the older page increasingly loses to a fresher one — even when the older page is objectively better sourced.
We first noticed it on a client comparison page that had been top-1 in AI Overviews for "best [category] platform" for eight months straight. It dropped to a supporting citation in late June with no content change on their side. What changed: two competitors published Q2-dated refreshes on their own comparison pages. That was enough.
“The citation-freshness penalty doesn't announce itself. Your traffic drifts down 15–20% over a quarter and you blame the season. It's not the season. It's that the model quietly stopped trusting your date.”
What the data shows
Aggregated across ~40 tracked prompt baskets in the second half of Q2:
- Pages with
dateModified< 90 days old: baseline citation rate. - Pages 90–365 days old with a visible changelog: −8% citation rate.
- Pages 365–540 days old, no visible refresh: −24% citation rate.
- Pages 540+ days old, no visible refresh: −41% citation rate.
The effect is strongest on ChatGPT Search and Google AI Overviews, meaningful on Perplexity, and barely detectable on Claude — which appears to weight authorship signals more than recency.
Why blanket "last updated" bumps don't work
The obvious hack — bump every dateModified to today across your whole site — is already discounted. LLM crawlers appear to cross-check the visible page content against the claimed date. If the content didn't meaningfully change but the date did, the freshness signal gets ignored (and, anecdotally, the domain seems to get a small trust discount).
What does work: a real, dated changelog entry that names what changed. Even one line ("Q2 2026: refreshed pricing data, added two competitor rows") is enough. It gives the model something to verify.
“Blanket date bumps are the AI-era equivalent of keyword stuffing. The models can tell, and the discount is worse than doing nothing.”
What to do next week
- Run a quick audit: which of your top-10 traffic pages have a
dateModifiedolder than 18 months? Those are your first refresh targets. - Refresh the content meaningfully — one new data point, one new competitor, one updated stat. Not a paragraph shuffle.
- Add a visible dated changelog line at the top or bottom of the page. This is the signal the models actually verify against.
- Bump
dateModifiedin JSON-LD and your sitemap<lastmod>in the same commit. See our note on schema for AI for the exact fields.
AI Visibility Watch is a weekly Friday issue from Crackle PR. Read the full series.
Founder & CEO, Crackle PR · Full bio
Parry Headrick is the founder and CEO of Crackle PR, a fully remote, senior-led B2B tech PR agency. With 20+ years in tech communications, he has counseled hundreds of VC-backed startups and growth-stage companies on media strategy, analyst relations, and AI-era brand positioning.
Glossary · Related Terms
- Google AI OverviewsGoogle's AI-generated answer block at the top of search results, synthesized from authoritative web sources.
- ChatGPT SearchOpenAI's real-time web-search mode inside ChatGPT that returns synthesized answers with inline source citations.
- Citation ShareThe percentage of AI-generated answers in your category that cite your brand as a source — the replacement metric for share of voice.
- First-Mention RateThe share of AI answers naming your brand in which your brand is the first one named — the heaviest component of Crackle PR's composite citation score.
- PerplexityAn AI answer engine popular with technical and research-driven B2B buyers; cites sources inline.
- On the RecordAnything you say can be quoted by the reporter with full attribution to you by name and title.
Read Next
- AI Visibility Watch
AI Visibility Watch #5 — July 3, 2026: Mid-Year Citation Audit, What Actually Changed in B2B GEO Since January
A mid-year audit of what shifted in B2B GEO since January — surface mix, citation density, and the three changes that moved the most pipeline.
Read “AI Visibility Watch #5 — July 3, 2026: Mid-Year Citation Audit, What Actually Changed in B2B GEO Since January” - AI Visibility Watch
AI Visibility Watch #4 — June 26, 2026: Claude Adds Web Citations & What It Means for B2B Brands
Anthropic enabled web citations in Claude by default. Early data on which sources Claude trusts and what B2B brands should do this week.
Read “AI Visibility Watch #4 — June 26, 2026: Claude Adds Web Citations & What It Means for B2B Brands” - AI Visibility Watch
AI Visibility Watch #3 — June 19, 2026: AI Overviews Hits the 'Best Of' Long Tail
Google AI Overviews now answers long-tail 'best [category]' queries — reshaping which B2B vendors buyers see first. What to do this week.
Read “AI Visibility Watch #3 — June 19, 2026: AI Overviews Hits the 'Best Of' Long Tail”