Citation Decay

Citation decay is the decline in how often AI search engines cite a page or domain after it has been cited regularly. It can happen while the page itself stays unchanged, because the answers around it change: newer pages compete for the same slots, the facts on the page go out of date, or the AI platform changes how it retrieves and selects sources.

It is the AI search counterpart of content decay, which describes falling rankings and traffic. The two often move together, but not always. A page can keep its Google ranking and still lose ChatGPT citations if ChatGPT's retrieval stops pulling it in, and a page can keep being cited by Perplexity long after its organic traffic faded.

There are three main causes, and each leaves a different pattern:

• Staleness and competition produce gradual decay on individual pages. Prices, statistics, product names and dates go out of date, and fresher pages on the same topic replace them in answers.
• Model and retrieval updates produce sudden, broad drops. Promptwatch's ChatGPT citation drop data shows ChatGPT citing about 27% fewer sources per response around the GPT-5.3 rollout on March 4, 2026, across every model in the product.
• Source-level shifts hit one domain or source type. On August 14, 2026, reddit.com's share of ChatGPT citations fell from about 4% to about 0.5% in a single day, and the share held by the largest DR 91 to 100 domains dropped at the same time.

Decay speed depends on the type of page. Promptwatch's ChatGPT content half-life study tracked pages ChatGPT cited on one day and checked which were still cited over the next 90 days. After 30 days, 78% of landing pages and 76% of documentation pages were still being cited. Product pages, reviews and how-to guides reached their half-life, the point where half had stopped being cited, after 74 to 77 days. News decayed fastest: almost a quarter of news pages were never cited again after the first day, and half were gone within 36 days.

Citation decay is measured per page over time. Useful indicators are the date a page was first cited, how long it took to reach its peak citation rate, and how many days after that peak its weekly citation rate halved. Promptwatch's Citation Trends view reports these as first cited, ramp up, peak citation and decay rate over a trailing 90-day window. Reading decay against platform-wide data tells you which cause you are facing: if every source dropped on the same day, it was a platform change; if only your page is fading, it is your content or a competitor.

The response depends on the cause. For gradual decay, run a refresh program for pages that earn citations: update facts, dates, prices and examples, add missing sub-topics, and keep the URL stable. Prioritize by format, refreshing news and product pages more often than landing pages and docs. For platform-driven drops, avoid rewriting good content in a panic; check whether competitors lost citations too, and spread visibility across more sources and engines so one change does not wipe it out. Track citation share and content freshness together so that decay is caught early rather than discovered in a quarterly report.