Content refresh case study: how to prove a refresh worked

We planned a data write-up. It became a measurement guide because our approved evidence is smaller than the title we had in mind.

Why is there no 50-post dataset on this page?

No approved Top SEO Experts case study covers a batch of refreshed posts for a home-service or any other local business. Our case studies report Search Console clicks and impressions for whole sites. They do not report before and after results for fifty individual rewrites.

The figures we can use sit on our case studies page, each with a period and source. None isolates refreshes from everything else that changed on the site. None reports leads or revenue, so treating them as refresh evidence would be a stretch.

A real “what refreshing 50 posts did” page needs these:

• The URL list.
• The date of each edit.
• Search Console exports before and after.
• A comparison group.

We use that checklist on other people’s write-ups too. See how to evaluate SEO case studies.

Every number in the worked example is hypothetical, and we label it. The rest is quoted from a Google page we opened on 22 October 2026, or stated as our view.

What does Google say about refreshing old content?

Google’s guidance supports updating pages that have become inaccurate. It warns against updating them only to look new. It publishes no refresh schedule and no rule that fresher pages rank higher, so a promised freshness boost goes beyond the documentation.

Keep three statements from Google next to your change log (accessed 22 October 2026). Each one limits what you can claim about a refresh.

• The people-first content guidance lists a warning sign. It asks: “Are you changing the date of pages to make them seem fresh when the content has not substantially changed?” The guidance also says informational content “should be factually accurate.”
• The publication dates page says to show a visible date with a clear label such as “Last updated.” It also says to keep visible and structured dates consistent, and never use future dates.
• The sitemap documentation says lastmod “should reflect the date and time of the last significant update to the page,” and that Google uses it only when it is consistently and verifiably accurate.

The thread is honesty about what changed. A corrected price range and a better answer earn a new date. Swapping “2025” for “2026” in the title does not.

Google’s Article documentation defines dateModified as “the date and time the article was most recently modified” and recommends a timezone. Our content refresh guide covers the workflow. This post is about proving the result.

Which posts are worth refreshing?

Refresh a post when it meets all three tests:

• It still matches a query people use.
• It has real impressions.
• It has one specific problem: stale facts, a thin answer or a title that loses clicks.

Leave well-performing posts alone. Merge or retire any post that competes with a better page.

This triage is our view, built from Search Console data you can pull yourself. The thresholds are starting points, not research findings.

What does Google’s debugging guide add?

Google’s traffic-debugging guide tells you to set the date range to the last 16 months. That shows whether a drop repeats yearly. The guide also says to read impressions and clicks together.

How do you use the table on your own posts?

Pull the last 16 months of Search Console data for each post. Match the post to the row that fits, then pick the default action for that row.

What does the table look like on two hypothetical posts?

Now apply the table to two hypothetical posts, with the arithmetic shown. Post A went from 3,000 impressions and 90 clicks to 3,100 impressions and 60 clicks over matched 90-day windows.

Impressions held, so the page still ranks. Something above it or in its title changed. The fix is a new title and opening answer.

Post B lost impressions and clicks together for four months, so it gets the body refresh. Same symptom, different edit.

When is merging the right move?

Merging is a different job from refreshing. Google’s guide to consolidating duplicate URLs says to use redirects when you want to get rid of existing duplicates. Its redirect guide says to use permanent redirects when you are sure the redirect won’t be reverted.

Our content audit guide shows how to find the overlaps.

What editorial rules do we apply when we refresh a post?

We apply eight rules, in this order. That way two editors refreshing two different posts make the same kind of decision. A bad rule can also be traced afterward.

1. Name the query first. Write the one question the post should answer and the Search Console queries it already gets. If you cannot, you are rewriting blind.
2. Verify every fact. Numbers, laws, features and prices each get a primary source and an access date. If you cannot confirm a claim, delete it.
3. Answer first. The first 40 to 60 words answer the query completely, with no link or pitch.
4. Keep what already works. The URL, the strongest headings and any passage earning impressions stay unless you have a reason.
5. Add what the top results lack. A decision rule, a worked example, a checklist or an honest limit. A longer page with the same information is not a refresh.
6. Cut before you add. Remove duplicated paragraphs, filler and sections that answer a different question.
7. Date it honestly. Update the visible date and dateModified only for substantial changes, with a change log entry for the same day.
8. Link both ways. Point the post at the service page it supports, and link to it from newer pages on the topic.

• Name the query firstWrite the one question the post should answer.
• Verify every factEach fact gets a primary source and an access date.
• Answer firstOpen with a complete answer and no link or pitch.
• Keep what already worksKeep the URL, strong headings and passages earning impressions.
• Add what the top results lackA decision rule, worked example, checklist or honest limit.
• Cut before you addRemove duplicates, filler and off-topic sections.
• Date it honestlyUpdate the date only for substantial changes.
• Link both waysLink to the service page and from newer pages.

If AI drafts any section, note Google’s spam policy on scaled content abuse. It describes many pages generated primarily to manipulate rankings. It says it applies however the content is made.

Our post on AI content at scale explains where we put the line.

What does a before and after look like?

A good before and after shows the edit, not only the metric. The example below is hypothetical: an invented post from an invented roofing company, written to show rules 3, 5 and 6. No real page or result sits behind it.

Before (opening of a hypothetical post titled “Roof replacement tips”):

Roofs have been protecting homes for thousands of years, and today’s homeowners have more options than ever. In this post we will look at some things to think about when deciding whether it is time for a new roof. They include materials, timing, and more.

After (same URL, new title “Roof repair or replacement: how to decide”):

Repair the roof if the damage is limited to one area and the rest of the surface is sound. Replace it when damage is spread across slopes, leaks keep returning after repairs, or your roofer finds the underlayment has failed. Get a written inspection before you decide either way.

What changed:

• The answer comes first and stands alone. A reader, or an AI answer engine quoting the page, gets a usable response in the first sentence.
• The history lesson is gone (rule 6), and the URL stays so links carry on.
• The next section would add a decision table and inspection questions (rule 5).

The after text makes no numeric claims on purpose. If the next paragraph quoted how long a roof lasts, that line would need a source and a date. Otherwise it would come out under rule 2.

A real version would show the old page, the new page and the Search Console line for that URL. We do not have that for a home-service client today.

How do you measure whether a refresh worked?

Compare the refreshed URLs to their own past and to untouched similar URLs over the same dates. Use Search Console clicks and impressions first, and calls or form fills second. A before and after with no control group cannot separate your edit from the market.

This is the method we would run for any batch, whether it holds 8 posts or 50, in this order:

1. Freeze a baseline. Export Search Console performance data by page for the 90 days before the edit. Google’s performance report help defines clicks as the number of times a user clicked your site from Google Search results.
2. Pick a control group. Choose posts of similar age, topic and traffic that you will not change. Aim for as many as you refresh, or at least 15 to 20.
3. Log every edit. One row per URL: date, what changed, which rule drove it, who edited.
4. Edit in batches. Do eight to twelve posts at a time and record the batch date. Then you see whether a rule works before you apply it everywhere.
5. Wait. Google’s SEO starter guide says changes might take effect in a few hours or take several months. We read a first result at 28 days and the real one at 90.
6. Compare matched windows. Same length, same weekdays, same season, year over year where the site is old enough.
7. Read impressions and clicks together. Impressions up and clicks flat points to a title or snippet problem. Both up is the pattern you want, and both down is a ranking or demand problem.
8. Connect to leads. Match landing pages to calls and form submissions. Our GA4 for SEO guide and Search Console performance report guide cover setup.

• Impressions up, clicks flatPoints to a title or snippet problem.
• Both upThe pattern you want.
• Both downA ranking or demand problem.

Paste this change log header into a spreadsheet and keep one row per refreshed URL:

url, edit_date, batch, rule_applied, what_changed, date_updated_on_page, editor, baseline_clicks_90d, post_clicks_28d, post_clicks_90d, group, notes

Keep “group” honest. If you refresh a control post halfway through, move it out of the control group and note the date.

What would a 50-post result look like?

Report a batch as a distribution plus a net lift against control, not a single total. In the hypothetical below, the refreshed group grows about 10 points faster than control. Yet about a quarter of the posts fall.

Everything in this section is hypothetical, with the arithmetic shown. It is not a Top SEO Experts result.

Imagine a service business with 100 comparable posts. Fifty are refreshed and fifty are held back as control. Search Console clicks over matched 90-day windows:

• Refreshed group15%
• Control group5.1%

Source: Hypothetical example

View the data as a table

Item	Value
Refreshed group	15%
Control group	5.1%

To read the table, compare each group with its own baseline first. Then subtract the control change from the refreshed change. What is left is the net lift.

Inside the refreshed group, also hypothetical:

• 18 posts up more than 20%.
• 20 posts within 20% either way.
• 12 posts down more than 20%.

• Up more than 20%18 posts
• Within 20% either way20 posts
• Down more than 20%12 posts

Source: Hypothetical example

View the data as a table

Item	Value
Up more than 20%	18 posts
Within 20% either way	20 posts
Down more than 20%	12 posts

What do the two big findings mean?

In this hypothetical, the control group rose 5.1%. A naive before and after would credit the refresh with 15%, when about a third of that came from the market.

A 15% average also hides 12 losers.

How do you read the 12 losers?

The useful work is reading those 12. Which rule did they share, and did the edit remove a passage that answered real queries?

We stop at clicks because we have no cost or lead data, and an invented dollar figure would look like a result. For the lead math on a real business, see our lead value math for home services.

Why do some refreshes not work?

Refreshes fail for reasons that have little to do with effort. Demand fell, the page’s intent changed, the edit removed an answer people used, or the window was too short.

For a post that lost ground after an edit, we check the causes in this order, cheapest first:

• Demand moved. If the control group fell too, the market moved, not your edit. Google’s debugging guide suggests Google Trends for this check.
• A reporting artifact. The same guide starts with the Data Anomalies page, because a logging error can look like a real drop.
• An algorithm or product change. Our Google algorithm updates post tracks what to check against your edit dates.
• The edit changed the page’s job. A post that ranked for “cost” and was rewritten as “how to choose” now answers a different question.
• Too little time. After two weeks you have no result yet. Our post on how long SEO takes explains the usual lag.

When the edit itself is the cause, restore the removed passage or revert the page. Reverting is not failure. It is the control group doing its job.

What does this data not show?

The hypothetical numbers show how to read a result, not what refreshes typically deliver. We have no dataset giving the average lift from a refresh for a service business. Anyone quoting one without a method and a date deserves scrutiny.

• No Top SEO Experts refresh dataset. We have not published one, so we cannot state our hit rate.
• No lead or revenue evidence. Clicks are a proxy. A refresh that adds clicks but no calls is a weak result.
• Small samples mislead. With eight posts, one page is a big share of the total. If one page holds half the clicks, its change moves the total by half as much.
• Other studies differ. Public refresh write-ups each use their own sample, definition and window. We have not verified their raw data, so we do not quote their figures.

Run the method, keep the change log, and in a quarter you will have your own evidence. Our original research content post covers publishing it. The service view is on our content SEO page.

Frequently asked questions

Next step

Before you refresh anything, write the one-question brief for each post. Our SEO content brief generator gives you a structure for the query, the answer and the sources. Then Get Your Organic Growth Plan.

Sources

• Google Search Central: Creating helpful, reliable, people-first content, accessed 22 Oct 2026
• Google Search Central: Publication dates in Google Search, accessed 22 Oct 2026
• Google Search Central: Build and submit a sitemap, accessed 22 Oct 2026
• Google Search Central: Article structured data, accessed 22 Oct 2026
• Google Search Central: Debugging drops in Google Search traffic, accessed 22 Oct 2026
• Google Search Central: Consolidate duplicate URLs, accessed 22 Oct 2026
• Google Search Central: Redirects and Google Search, accessed 22 Oct 2026
• Google Search Central: Spam policies for Google web search, accessed 22 Oct 2026
• Google Search Central: SEO Starter Guide, accessed 22 Oct 2026
• Google Search Console Help: Performance report, accessed 22 Oct 2026