A decaying page is not one with fewer visitors. It still ranks for its old term, still gets clicks, and has quietly stopped being quoted. AI answers credit a competitor's page for the same question now, and because the click never dropped, nobody inside the practice noticed anything had changed.
What Content Decay Looks Like From The Inside
Ranking and citation are different measurements of the same page. Ranking asks whether you appear in a list. Citation asks whether you appear in an answer. A page can hold a position for years and never be named in a paragraph, because the question that produced the answer no longer matches what the page says.
Ranking Is Not Citation
The classic decay profile looks like this: search impressions flat, organic clicks flat, mentions in AI answers falling away. If your reporting stops at traffic, decay is invisible, because traffic was never the thing that changed.
That is why the measurement has to be about mentions on the questions your page targets, not impressions on the keywords it targets. At drsmiles.ai the measurement runs across ChatGPT, Gemini, Perplexity and Claude, with the same questions asked in the same way, so a decline in citations shows up as a number rather than a suspicion.
The Four Causes Of Decay
Almost every decayed page has one of four problems, and the table below shows how each one announces itself.
Swipe the table sideways to see every column.
| Cause | What It Looks Like | Quick Check |
|---|---|---|
| Stale facts | The address, hours, clinician names or service list have moved on since publication | Read the first two hundred words and compare every movable detail against what the practice offers today |
| Broken links and orphans | No page on the site links to it, and it links nowhere useful | Run an internal link report and list pages with zero inbound links |
| Wrong format for the question | The answer is buried below paragraphs of context instead of stated up front | Read only the first paragraph and ask whether a machine could quote it |
| A fresher rival source | A competitor published the same material more recently and now gets named instead | Read the source list an AI answer returns and look for their page where yours used to be |
Stale facts are the quietest of the four. The page says something that was true when it was written: the practice relocated, the clinical team changed, the appointment policy changed, the building was refurbished. Engines do not check your page before quoting it, and neither do patients.
Broken links are the second cause. A page with no internal links pointing to it, and none pointing out, is a page the site structure has forgotten. It may still be found, it may still rank, and it is rarely trusted enough to be quoted.
Format is the third. An article written in 2019 that opens with four paragraphs of background will lose to a page that answers in the first sentence, even when the older page has more detail and more links pointing at it.
The fourth cause is simply a competitor with a fresher source. When somebody republishes your material with a current date while your page still shows its original year, the newer version becomes the easier citation and the older one becomes the fallback.
Finding The Corpses In Your Library
Three searches turn decay from a feeling into a list you can work through.
The Three Searches
- Compare citation lists. For each question you care about, read the sources an AI answer actually names. A page of yours missing from a list where a competitor's page appears is a page you have already lost, whatever your ranking report says.
- Run an orphan check. List the pages with no internal links pointing in. These are the bodies: reachable, sometimes visited, structurally abandoned.
- Fact-check by hand. Take the pages that used to perform best and check the five details most likely to have moved: address, opening hours, clinician names, services offered and dates.
The first two checks are mechanical once the tools are in place. The third takes longer and finds more, because a page can keep its links and its ranking while describing a practice that no longer exists in that form.
Two of the three checks can be started with the visibility snapshot on the home page, which asks ten questions across three engines in about an hour and shows where you are missing from an answer. The full comparison across four engines and the source lists behind each answer are in the sample report.
The Refresh Protocol
Refresh in this order. Each step depends on the one before it, which is why working through them out of sequence wastes the effort.
- Correct the facts. Reason first: a page that quotes a wrong number will keep quoting it after the refresh, leaving you to tidy the formatting of a false statement.
- Re-anchor the answer. Reason second: put the direct answer into the first two sentences, where an engine can lift it without rewriting the meaning.
- Restore the links. Reason third: link the page into its cluster and back out to the pillar, so the site structure tells engines what the page is for.
- Update the date honestly. Reason fourth: change the date only when you have changed the content, and add one line saying what changed.
- Re-submit the URL. Reason last: this tells crawlers the page has moved on. It is a notice rather than a request for a position.
The order also protects you from the classic mistake, which is polishing the wording while leaving the broken facts in place.
How To Refresh Without Starting Over
Most decayed pages do not need a rewrite. They need the current facts, the answer moved up, and the links back.
- Change the facts that are wrong before you change anything else
- Move the direct answer into the first two sentences and leave the rest of the page where it is
- Keep the URL, the title and every external link you have earned
- Add three internal links: one to the pillar, one to a sibling page, one to a contact or booking page
- Change the visible date only if the content genuinely changed, and note what changed in a sentence
- Leave the original conclusion alone while it is still true
A Thirty-minute Rescue
When you have half an hour, use the short version: five minutes to fact-check, ten to rewrite the opening as a direct answer, ten to add or repair the internal links, and five to update the date line and re-submit the URL. That recovers most of the value on a page whose substance is still correct.
It does not rescue a page whose central claim has changed, and it does not replace the cluster work described in building topical authority for a dental practice. A page with nothing unique left in it belongs merged into a stronger page rather than polished on its own.
Preventing Decay
Decay is a process problem, so prevention is a process: a review rhythm, a date you keep, and a name against every page.
- Review the library one quarter at a time, with the orphan report open beside you
- Give every page an owner who works there and a review date you actually maintain
- Re-check the five movable facts whenever you touch a page rather than on a fixed schedule
- Keep a current list of the pages that AI answers name, so you know what is working
- Feed every question you keep losing into the next piece of content
Cadence is the other half of prevention, because a page with a fixed review rhythm is a page that gets reviewed. how often a practice should publish covers the rhythm question, and what E-E-A-T means for dental content covers the reviewer question, since every page needs a named person who can answer for it. Where pages are reviewed, argued over and kept current is where the AEO service earns its place.