Refresh, Rewrite, Prune, or Redirect: How to Decide What to Do With Each Page

You open the traffic report, see a page sliding, and freeze. Update it? Start over? Delete it? Send it somewhere else? This guide hands you a content decision framework for one page at a time, so every underperforming page decision ends in a defensible action instead of a guess. By the end you will know how to read the evidence, pick the least destructive fix, and write down why.

If this feels like a lot, that is normal. You only need to work one page at a time.

Know what each of the four actions means

Most teams argue about the wrong thing because the words are fuzzy. Let's fix that first.

• Refresh: the topic and the reader are still right. The page is just stale. You improve it and usually keep the URL.
• Rewrite: the topic still matters, but the page was weak from the start, aimed at the wrong intent, or built in the wrong format. You rebuild the answer.
• Redirect: this URL should not stand alone anymore, and a close, genuinely useful replacement exists. You send readers there.
• Prune: the page has no meaningful value left, cannot be made accurate or useful, and has no close replacement. You take it out of the active set.
• Something else entirely: the real problem is seasonality, demand, indexing, authority, tracking, a recent bad edit, or the conversion path. No content action fixes any of those.

There is a fifth move hiding in that list: consolidate. When two pages fight over the same intent, you merge the useful parts into the stronger one and redirect the weaker URL. Consolidation is a content decision and a URL decision at the same time.

Here is the rule that carries the whole guide. Match the action to the diagnosis. Refreshing a page that needs consolidation wastes a week. Redirecting a page with no close replacement frustrates readers. Deleting a page with real backlinks throws away an asset. Rewriting a page that only needed current numbers is work you did not have to do.

Notice what is missing. Nowhere does this say "always refresh." Least destructive does not mean least effort. It means do not throw away a useful URL when a focused change solves the problem, and do not keep a page alive when it cannot do a useful job.

Step 1: Write down the page's job and its baseline

Before anyone proposes an action, describe the page. This takes ten minutes and saves you from arguing in circles later.

Write down:

• Who the page is for and what problem it solves.
• Its primary topic, target query cluster, and search intent.
• Its funnel stage and business role: awareness, consideration, decision, support, or trust.
• The conversion or next action it is meant to support.
• Its publication date and the last time someone changed it substantively.
• Any unique brand, product, research, compliance, or customer value that traffic data will never show.
• Its URL, its internal-link role, known backlinks, and any page covering a similar topic.

Then pull the numbers. In Search Console, compare clicks, impressions, CTR, and average position across the last 3, 6, and 12 months, plus the same period a year earlier if the topic is seasonal. A 28-day window gives you a stable signal. A 7 or 14 day window is an early alert, not a verdict. In analytics, look at organic landing-page sessions, engaged sessions, and conversions.

Read the metric definitions carefully, because a lot of bad audit calls come from misreading them. An impression is an appearance in results, not a visit. A click is a visit from a search result. CTR is clicks divided by impressions, and it moves with your snippet, the layout of the results page, and intent, not only with your body copy. Average position is an aggregate across queries, devices, and locations, so read it as a trend line rather than a rank.

Done when: your audit row describes the page's job, its baseline, its business value, and the windows you compared, all before anyone says the word "rewrite."

Where people go wrong: they call a page a failure because its absolute traffic is small. A quiet decision-stage page that closes deals can be worth more than a busy awareness post.

Step 2: Confirm the decline is real

A single bad week is an alert. It is not evidence. Normalize the signal before you touch the page.

Work through five checks.

1. Compare matched periods. Account for seasonality, holidays, day-of-week patterns, and small samples.
2. Check the scope. Is this one page, a query cluster, one device group, one country, a whole template, or the entire site? A site-wide problem is not a page problem.
3. Rule out technical explanations at a high level: indexing, availability, redirects, analytics tagging. If something looks broken, hand it to the person who owns it and stop there.
4. Check whether the page changed just before the drop. Compare it with the previous version. If an edit removed something useful, restoring the stronger version may beat any new work.
5. Check links and difficulty. If the page has lost referring domains, or the query is simply harder than your site's authority, content work will not close that gap.

Practitioner frameworks use working thresholds to decide what deserves a look. Treat them as investigation flags, never as automatic actions:

• More than a 20 percent organic-traffic decline over 90 days.
• More than a 20 percent year-over-year decline, useful as a quarterly flag.
• Three consecutive 28-day periods with at least a 20 percent click decline and a drop of two positions or more, which is one framework's definition of structural decay.
• A 30 percent click decline against a 28-day baseline for two straight weeks, used elsewhere as a discovery alert.
• A 25 percent CTR decline while average position moves less than half a point, which points at the snippet or a search feature rather than the page body.
• A drop of two positions or more on queries with at least 100 impressions in 28 days.
• A 30 percent impression decline while position holds, which often means demand or indexing changed.
• A 20 percent drop in average engagement time while traffic holds steady.
• A 25 percent conversion-rate decline while sessions hold steady, which usually means the offer or the path, not the words.
• Six months or more with no new referring domains.

Do not stack these into a scoring formula. They are guardrails for a conversation, not a verdict. Low-volume pages need absolute cutoffs too, because 40 percent of nine clicks is noise.

Pro tip: keep the page's current numbers, the comparison window, the last substantive edit, and the suspected cause in the same audit row. That one habit stops your team from confusing a seasonal dip, a bad edit, and genuine decay.

Done when: you can say whether the change is persistent, comparable, and probably about the content, or whether this page belongs in a technical, demand, authority, UX, or measurement review instead.

Where people go wrong: they apply a 20 percent rule mechanically, compare a quiet month with a busy one, or refresh a page whose real problem was a broken template.

Step 3: Diagnose the failure pattern

Now read the shape of the decline. The pattern tells you where to look next, and that is the difference between a fix and a guess.

Then go look at the live results for the page's leading queries. What format wins now? Guide, comparison, list, product page, video, forum thread? A change in the dominant format is a rewrite signal. A competitor with a fresher page is a warning, not proof that more words will win.

Track your AI health as a separate line. A page can hold its rankings and quietly stop being cited by answer engines, or keep its citations while Google traffic slides. Those two situations need different work, so measuring them as one number hides the problem.

This is where DeepSmith's AI Visibility helps. It reports mention rate, citation rate, and share of voice for the prompts you track, breaks results down by platform, shows which competitors get cited instead of you, and shows exactly which of your pages earn citations. Use it next to Search Console and analytics, not in place of them. Plan coverage matters here: Pro tracks ChatGPT, Grow adds Perplexity, Scale adds Gemini, and Enterprise or Custom covers all ten engines.

Done when: the audit row names one primary suspected cause, and a secondary one if the evidence is split. "The page is down" is not a cause.

Where people go wrong: they read every position loss as a content problem, every CTR loss as a rewrite, and every missing AI citation as proof the page is worthless.

Step 4: Test intent and content quality

This is the refresh vs rewrite step, and it comes down to one question. Is the foundation sound?

Ask these about the page as it stands today:

• Does it still answer the query and the audience need it was built for?
• Has intent shifted, from informational to commercial, or from guide to comparison?
• Does it offer original information, analysis, examples, or a point of view, or does it summarize other people's pages?
• Is the explanation complete enough that the reader does not immediately search again?
• Is the title honest and descriptive rather than exaggerated?
• Are there factual errors, weak sourcing, or no visible expertise?
• Was it written for a real reader, or for a search engine?
• Is it padded to hit a word count?
• Would it help someone who arrived directly, with no keyword involved?

Choose a refresh when the answers are still yes and the gaps are specific: stale statistics, old screenshots, dead links, dated terminology, a muddled structure, a weak call to action. The promise of the page is intact. You are bringing its substance current.

Choose a rewrite when the topic is worth owning but the page cannot get there from here. Intent shifted. The format is wrong. The treatment is thin or derivative. Several sections fail on their own. A rewrite has to produce real information gain, not more paragraphs.

A rewrite is not the braver, better choice. It costs more and it risks more, because it can strip out language, rankings, and links that were quietly working. Keep the URL when topic and intent still line up and the URL has equity worth protecting. Do not keep a URL if that means misleading the reader about what the page now is.

Common mistake: using word count as the rewrite test. The real tests are information gain, accuracy, completeness, usefulness, and intent fit. No preferred length decides this.

Done when: you can finish one of two sentences out loud. "This page is still about the right thing for the right reader, it just needs a refresh." Or, "The topic is right, but this page needs a rewrite because ..."

Where people go wrong: they treat freshness as a date change. Update the substance first, then move the date. A date that moves without the content moving is a promise you did not keep.

Step 5: Check equity, overlap, and whether a replacement exists

Before you decide to prune or redirect anything, find out what you would be throwing away and where readers would land.

Record:

• Backlinks and referring domains, especially links pointing at a statistic, a quote, or original research.
• Historical peak traffic, current traffic, rankings, conversions, and the page's internal-link role.
• Brand, product, support, legal, or historical value that analytics understates.
• Overlapping pages, with their query sets, intent, traffic, and authority.
• The strongest possible successor page, if there is one.
• Whether that successor actually satisfies the old page's reader.
• Whether the page is obsolete, expired, inaccurate, or risky to keep live.
• Whether the page earns AI citations today, or whether a competitor page has taken that role.

When pages compete, consolidate. Several thin pages chasing the same intent should not each get their own refresh. Pick the survivor on intent fit, traffic, links, authority, and business value. Absorb the useful material from the others. Then redirect the redundant URLs to the survivor.

Redirect when there is a close replacement. A redirect tells readers and search systems that this page's job now lives somewhere else. Make it permanent when you do not expect to bring the page back.

Prune when there is not. No close replacement, no meaningful equity, no legal or support or brand reason to keep it: that is a pruning candidate. Do not point a redirect at your home page just to avoid a removed URL. An unrelated destination does not become relevant because you sent traffic to it.

If the page has real backlinks but the topic no longer fits your strategy, look harder for a relevant successor. Routing that equity somewhere useful beats deleting it.

DeepSmith's Content Map is useful at this step. It crawls your pages and your competitors' pages into one topic and funnel-stage taxonomy, then shows coverage gaps, untapped topics, and how deep each topic goes. Overlap becomes visible: whether a topic is thin, doubled up, or missing its decision-stage page. Treat that as evidence for your decision, not as a verdict on whether a page should exist.

Done when: the audit row has an equity assessment, an overlap assessment, and a one-word replacement answer: close, none, or not applicable.

Where people go wrong: they redirect retired pages to the home page, delete pages without checking links and conversions, or merge two pages that share keywords but serve different readers.

Step 6: Choose the action with the decision tree

Now run the branches in order. Stop at the first one that fits.

1. Is the problem content-related and persistent? If no, route it to whoever owns the technical, demand, authority, UX, or measurement side. No page action.
2. Do the topic and intent still matter? If no, ask whether a close replacement exists. Yes means redirect. No means prune, after your business, legal, link, and reader checks.
3. Is another page competing for the same intent? If yes, consolidate into the strongest survivor and redirect the rest.
4. Is the foundation sound, with mostly stale or incomplete parts? If yes, refresh.
5. Is the topic valuable but the foundation, intent fit, format, originality, or structure fundamentally weak? If yes, rewrite.
6. Does the page have no meaningful value and no close replacement? If yes, prune.
7. Is the evidence mixed? Leave it in investigation. A permanent decision on weak data is the expensive kind of mistake.

The matrix version, for your audit sheet:

Common mistake: turning this into "update or delete." Those are two of at least six real options. Rewrite, consolidate, redirect, restore, and "not yet" are all on the table, and the binary version pushes you into the wrong one.

Done when: a second reviewer can read your evidence and land on the same action without asking what "underperforming" meant. Every underperforming page decision should be reproducible by someone who was not in the room.

Run the same content decision framework on the next page and the branches start to feel automatic.

Where people go wrong: they decide from traffic alone. Traffic is one input out of eight.

Step 7: Rank the backlog and record the decision

You now have a decision. Turn it into a queue, because deciding what to do with old content only pays off when the work actually gets done.

Rank on impact against effort. The factors that matter most:

• Business relevance and how close the page sits to revenue.
• Historical peak traffic and the size of the current loss.
• Conversions and assisted conversions.
• Keyword value and whether you can realistically compete for it.
• Backlink equity and the quality of those referring domains.
• AI citation and mention opportunity.
• The effort and risk of the action you chose.
• Whether a replacement already exists.
• Compliance, brand, customer, and support consequences.

High impact and low effort go first. Pages that score high on business relevance, past performance, and realistic competitiveness get fixed before anything else. Pages low on all three are your pruning shortlist. A high-potential page with a big effort attached gets planned deliberately, not squeezed in beside the quick wins.

For every page, record the evidence and the window, the suspected cause, the chosen action, the alternatives you rejected, what happens to the URL, the equity and replacement assessment, the owner, the priority, your confidence, and the metric that will tell you whether it worked.

That "rejected alternatives" line is the one people skip, and it is the one that saves you in six months when someone asks why a page is gone.

Set a rhythm you can keep. Short alerts catch fresh drops. A monthly scan catches a page before a quarter passes. A formal quarterly decay audit covers the library. An annual deep review covers your highest-value pages. After you act, recheck with a comparable window. Thirty days is a reasonable first look, though a low-traffic page or a big structural change needs longer.

Done when: the backlog is ordered, every page has one defensible action or an explicit "investigate another cause," and the reasoning is written down.

Where people go wrong: they treat the action as the finish line, never record the baseline, or borrow an uplift number from someone's case study and expect it on their own page.

What to do next

Deciding what to do with old content gets easier with reps, so start small. Take one page. Just one. Run it through the seven steps and put the row in a spreadsheet. Then do the next one. You will spot your own patterns fast, and the fourth page will take a quarter of the time the first one did.

Momentum matters more than a perfect audit template.

If the AI-search half of this is the part you cannot see today, that is the gap worth closing first. DeepSmith tracks which prompts mention you, which of your pages get cited, and which competitors are taking those citations, then maps your library and your competitors' libraries onto one topic taxonomy so overlap and gaps stop being guesswork. You can start a 7-day free trial and work with real data on your own pages before you pay.