No, you should not delete old content just because it is old — but yes, most small business blogs and resource sections are carrying pages that are actively hurting their visibility in AI search, and the fix is usually to delete, merge, or redirect them rather than leave them sitting there. The hard part is telling the two apart, which is what this framework is for.
We started taking pruning seriously for a different reason than the old SEO advice to “clean up thin content.” With AI answer engines — ChatGPT, Perplexity, Google’s AI Overviews — a site with three overlapping, half-finished pages on the same topic does not just waste crawl budget. It actively confuses the model about which page represents your actual, current answer, and it dilutes the specific, quotable signal that gets a page cited in the first place.
Why Pruning Matters More With AI Crawlers Than It Did With Google Alone
Google has spent two decades getting reasonably good at picking the “best” version of a topic out of several competing pages on your own site. AI answer engines are newer at this, and our experience pulling and reviewing AI citations for client sites suggests they are more likely to either cite an outdated version by mistake or simply skip a topic where your own site sends mixed signals across multiple pages.
There is also a more mechanical issue: AI crawlers work within a budget, just as Googlebot does, and a bloated site with hundreds of thin, overlapping posts spreads that budget thinner. A smaller number of comprehensive, clearly differentiated pages is easier for both a crawler and a model to make sense of.
The Four Buckets
Every page under review lands in one of four buckets. We run this exercise roughly once a quarter for client sites with active blogs.
Delete
Reserve this for pages with no traffic, no backlinks, no internal links pointing to them, and no topic relevance left — think an old “2019 trends” post, or a product announcement for something you no longer sell. Deleting these removes noise without losing anything worth keeping.
Merge
The most common bucket for agency and SME blogs. If you have three or four posts circling the same core question from slightly different angles — each thin, none comprehensive — combine the useful parts into one strong page and redirect the others to it. This is usually where the biggest visibility gains come from, because it turns three mediocre-signal pages into one page with a clear, citable answer.
Redirect (no merge)
For pages that are genuinely obsolete but map cleanly onto a newer, better page — an old “how to set up Facebook Pixel” post, for example, once you have a proper Conversions API guide — a straight 301 redirect is enough. No content merge needed, just point the URL and its accumulated links at the page that replaced it.
Keep and Refresh
Pages that still get traffic or citations but have gone stale — outdated statistics, a screenshot from an old platform interface, pricing that changed. These stay, but need an update pass rather than a rewrite.
How to Decide Which Bucket a Page Belongs In
We look at five signals together, not any single one in isolation:
| Signal | What it tells you |
|---|---|
| Organic traffic, last 12 months | Near-zero traffic plus no backlinks usually means delete or merge, not refresh. |
| Referring domains / backlinks | Even one or two solid backlinks is a strong reason to redirect rather than delete outright — you keep that link equity. |
| Internal links pointing to it | A page with many internal links pointing in is probably load-bearing for your site structure, even with low traffic. Investigate before removing. |
| Topic overlap with other pages | Two or more pages targeting the same core question is the clearest merge signal there is. |
| Factual freshness | Outdated numbers, pricing, or screenshots push a page toward refresh even if traffic is healthy. |
What About Pages That Rank But Never Get Cited by AI Answers?
This is a specific case worth calling out separately. A page can hold a decent Google ranking and still never show up in ChatGPT or AI Overview answers for the same query, usually because it buries the actual answer under throat-clearing intro paragraphs instead of stating it plainly near the top. Before you prune a page like this, try a lighter fix first: rewrite the opening two or three sentences to directly answer the core question, add a clear H2 structured as a question, and see whether citation improves over the following few weeks. Pruning is for pages with a structural or duplication problem — not every under-cited page needs to be merged away, some just need a sharper opening.
The Merge Process, Step by Step
- Group candidate pages by shared core keyword or question intent, not just similar titles — two pages can have different titles and still be answering the same query.
- Pick the strongest page as the target: usually the one with the most backlinks or the best current ranking, not necessarily the newest.
- Pull the genuinely useful, non-overlapping material from the other pages into the target — a specific example, a data point, a section structure that worked better.
- 301 redirect the merged-away URLs to the target page.
- Update internal links across the site that pointed to the old URLs, rather than relying on the redirect alone.
- Resubmit the target URL in Google Search Console and note the date, so you can track whether visibility and citation improve over the following weeks.
How Often Should You Run This?
Quarterly works well for an active blog publishing multiple times a week; twice a year is usually enough for a site publishing a handful of posts a month. The trigger that matters more than the calendar is volume: once you have enough posts that you are not sure yourself whether you have already covered a topic, it is time for a pass. Our own content audit playbook covers the broader underperforming-page review this pruning exercise sits inside of, if you want the fuller process.
What We Would Never Delete
A few categories are worth protecting even when the traffic numbers look weak:
- Pages with backlinks from sites you would struggle to get a link from again — redirect these, never delete outright.
- Pages that are part of a topic cluster supporting a pillar page’s authority, even if they individually get little direct traffic. We wrote about why this structural signal matters in our piece on topical authority.
- Recently published pages — give new content at least three to six months before judging it on traffic; AI citation in particular tends to lag actual publication by weeks.
Keeping This From Becoming a Recurring Problem
The real fix is upstream of pruning: a content calendar that checks for topic overlap before a new post is briefed, rather than one that discovers it a year later during an audit. It is also worth revisiting how your site’s internal linking structure handles related topics, since a lot of the pages we end up merging were never clearly connected to each other in the first place, which is part of why they existed as separate, competing pages rather than one page found and expanded twice.
The Bottom Line
Pruning is not about shrinking your site for its own sake. It is about making sure every page you keep is the clearest, most current answer to the question it is trying to answer — for a person, for Google, and increasingly for a model deciding which page to quote. Run the four-bucket exercise against your weakest quarter of posts first; that is usually where the clearest merge and delete candidates are sitting.