AI Citation Decay: Why Your Best Content Stops Getting Cited (And the Refresh Cadence That Works)

Your 2024 guidepost is already dead

We had a client who published a solid GEO guide in January. It got cited by ChatGPT within two weeks. By March, the citations had halved. By May, they were gone entirely. The content hadn't changed. The competition had.

That's citation decay, and it breaks the most basic assumption behind traditional SEO: that good content compounds over time. In AI search, it doesn't. Your best work has an expiration date, and that date is closer than you think.

The numbers back this up. A study by Scrunch and Stacker, analyzing 3.5 million citation events across ChatGPT, Google AI Overviews, Perplexity, Gemini, and Claude from September 2025 to March 2026, found a median citation half-life of 4.5 weeks. Half of all AI citations come from content less than 13 weeks old.

Why AI citations decay faster than search rankings

Traditional SEO decay is glacial. A well-built page can hold rankings for years, slowly losing ground as competitors earn more backlinks or publish more comprehensive content. The decay curve is gentle and forgiving.

AI citation decay works on a different timeline. Three reasons:

Retrieval systems re-rank on every query. Every time someone asks ChatGPT or Perplexity a question, the system re-evaluates which sources to cite. There's no "position" that sticks. The model scores sources against the specific query, and recency acts as a cheap proxy for accuracy. A fresher page routinely displaces an older one that might be equally good.

Citation pools are smaller. Google shows ten organic results. AI engines cite two to five sources. If you're not in that tiny pool, you're invisible. And the pool refreshes constantly.

No backlink moat. In traditional SEO, a page with 200 backlinks is hard to dislodge. In AI search, backlinks matter less than content structure, factual accuracy, and recency. A brand-new page from a less authoritative domain can replace yours if it answers the question more directly.

Decay rates by platform

Not all AI engines burn through citations at the same speed. The differences matter for how you prioritize your refresh calendar.

ChatGPT is the fastest chaser. Median citation half-life sits around 3.4 weeks. ChatGPT skews heavily toward encyclopedic and editorial sources. Wikipedia accounts for nearly 48% of its top cited domains. Freshness matters, but structure matters more. Paragraphs that start with direct definitions ("X is...") get extracted and cited more reliably than narrative intros.

Google AI Overviews land in the middle at roughly 4.3 weeks. Google's own data confirms AI Overviews now appear for about 48% of tracked queries. The platform pulls citations with a bias toward existing top-ranking content, but the new Authority Signals update (rolling out since September 5) is shifting that. Pages with demonstrated, verifiable expertise are being rewarded. Thin content optimized only for structure is losing ground.

Perplexity holds longest at about 5.7 weeks, but it's also the pickiest about recency. Perplexity strongly favors content published or updated within the last 90 days. It also leans heavily on Reddit and community sources (46.7% of top citations). If your content is more than three months old without a refresh, Perplexity will replace you with someone who published last week.

Decay rates by content type

The type of content you publish determines how fast citations fade.

News and trending content decays in days. If you publish a reaction piece to a Google algorithm update, expect citations to peak within 48 hours and drop sharply within a week. This content has a short shelf life by nature.

How-to and tutorial content holds for three to six months with periodic updates. But "periodic" means something different in AI search than in SEO. Updating a how-to guide every six months used to be fine. Now, if the topic is competitive, you need updates every six to eight weeks to maintain citations.

Technical reference content (specs, benchmarks, data tables) decays more slowly, around three to six months, because the information is harder to replicate. But once someone publishes fresher data, you're out.

Brand pages and product documentation decay slowest of all, often holding citations for six months or longer. But they're also the least likely to be cited in the first place. AI engines prefer third-party editorial sources over brand-owned content.

The refresh calendar we actually use

After tracking citation half-lives across dozens of client pages, we built a refresh cadence that matches the decay curve instead of guessing.

Weekly (news and trending): Refresh within 48 hours of publication. Add new data points, update timestamps, and refresh the opening paragraph with current context. This applies to reaction pieces, news commentary, and time-sensitive analysis.

Every 4 to 6 weeks (competitive how-to and comparison content): This is where most teams fall behind. If you published a "best tools for X" or "how to do Y" article, mark your calendar for four weeks out. By then, half your citations may be gone. The refresh doesn't need to be a rewrite. Update statistics, add a new section addressing recent changes, refresh the date, and ensure the structure still matches how AI engines extract passages.

Every 8 to 12 weeks (educational and framework content): Longer-form guides and framework pieces have more staying power. But you still need to touch them at least quarterly. Check that cited sources are still live, that recommendations still hold, and that no competitor has published a more comprehensive version.

Every 16+ weeks (reference and brand content): Technical documentation and brand pages need the least frequent updates. But "least frequent" still means at least twice a year. Verify that specs are current, that linked resources haven't moved, and that the content still answers the questions AI engines are asking.

How to detect decay before it costs you visibility

Most teams find out about citation decay when traffic drops. By then, you're already behind. Here's how to catch it earlier.

Track citation presence manually. Pick your ten most important pages. Once a week, search for the same questions those pages answer in ChatGPT, Perplexity, and Google AI Overviews. Note which pages get cited. When half your tracked URLs stop appearing, you've hit your half-life. It sounds tedious, but it tells you something no dashboard does yet.

Watch Search Console CTR by query, not just position. A page ranking well but sitting on a low CTR is a sign AI Overviews are intercepting clicks before they reach you. This is especially common on informational queries right now.

Monitor competitor freshness. If a competitor publishes a newer version of content you rank for, your citation clock is ticking. AI engines don't care about your seniority. They care about which answer is most current and most relevant to the specific query.

Publishing and forgetting is dead

Publishing and forgetting is no longer viable. In traditional SEO, you could publish a great piece and ride it for years. In AI search, every piece of content you publish is a commitment to maintain it.

This doesn't mean you need to rewrite everything monthly. It means you need a system. Track what's getting cited. Know your half-life by content type. Build refresh triggers into your editorial calendar, not as an afterthought but as a standing task.

The brands that figure this out will keep their citations while competitors fight for scraps. The ones that don't will keep wondering why their best content stopped working.

Preguntas frecuentes

¿Qué es la vida media de citaciones en búsqueda IA?

La vida media de citaciones es el tiempo que tarda la mitad de las citaciones de tu contenido en ser reemplazada por fuentes más recientes. Entre ChatGPT, Perplexity y Google AI Overviews, la mediana es de unas 4,5 semanas. Después de ese punto, la mitad de tus citaciones desaparecen.

¿Con qué frecuencia debo actualizar el contenido para mantener las citaciones de IA?

Depende del tipo de contenido y la plataforma. El contenido de noticias y tendencias necesita actualizaciones semanales. Las guías paso a paso deben renovarse cada 6 a 8 semanas. Las páginas de referencia técnica y las páginas de marca pueden aguantar más, unas 12 a 16 semanas. ChatGPT exige las actualizaciones más frecuentes (aproximadamente cada 3 a 4 semanas), mientras que Perplexity retiene las citaciones más tiempo, unas 6 semanas.

¿La decadencia de citaciones afecta por igual a todas las plataformas de búsqueda IA?

No. ChatGPT renueva las más rápido, con una vida media de citaciones de unas 3,4 semanas. Google AI Overviews está en torno a 4,3 semanas. Perplexity aguanta más, unas 5,7 semanas. La diferencia depende de cómo cada plataforma pondera la actualidad frente a la autoridad.