Updating content without losing rankings

First of all: work out why it dropped

The two jobs carry opposite risks. A bad update adds weight without changing anything. A rewrite on a healthy page throws away the ranking you meant to defend. Which is why the most expensive mistake is updating a page that does not have a content problem: if the drop comes from somewhere else, rewriting does not fix it and puts a ranking that was standing up back into play.

Google lists six families of causes for a drop in search traffic: algorithm updates, technical issues, security issues, spam issues (algorithmic or with a manual action), seasonality and changing interests, and site moves and migrations. Only one of them is cured by rewriting.

Two tools are enough to place the case. The Search Console performance report, set to sixteen months — the range Google recommends precisely so a drop can be put in context and checked for recurrence, comparing against the same period the year before. And Google Trends, to see whether the drop is yours or the whole sector's.

The case that confuses most: the core update

If the drop coincides with a core algorithm update the rules change, and this is the distinction that is almost always missing.

Google is clear that core updates do not target specific sites or pages, and that a page which falls is not necessarily poor: the image it uses is a restaurant ranking, where the ones that slip have not got worse — others have moved into the top twenty.

Three practical instructions follow, all of them Google's:

1. Wait. Before analysing anything, wait at least a full week after the update finishes rolling out.
2. Do not overreact. For a small drop, from second to fourth position, Google says explicitly that no drastic action is needed. A drop that takes a page from fourth to twenty-ninth does call for a deeper assessment — one Google frames around the site as a whole, not only the page that fell.
3. Do not expect speed or guarantees. Some effects show in days, others take months, and sometimes you have to wait for the next update. Google writes that it cannot guarantee changes will have a notable impact on results.

When updating is worth it

The clearest signal is a page sliding slowly: average position worsening on phrases it used to hold, with the drop spread over months rather than landing on one day. The question is still live, the answer is no longer the best available. The second signal is a page ranking on phrases it never aimed at: the question has widened and the page only covers it side-on.

One case belongs apart from the others. If impressions hold steady and only clicks fall, the page is still showing but people are choosing someone else. Before rewriting, look at the title, the snippet, and the elements that have appeared on the results page and are absorbing the click. You touch the content only after ruling those three out.

Not every page that drops should be updated. A page losing traffic because the question has disappeared should be retired, not rewritten, and telling the two apart is the step most often skipped.

Signals that justify an update

• Average position worsening slowly on phrases the page used to hold.
• Rankings on neighbouring phrases the page covers badly.
• Sections citing data or tools that are no longer current.
• A page answering a question that has since changed intent.

Signals that do not

• A drop that lines up with the same period last year.
• A drop affecting the whole sector.
• A drop of a few places right after a core update.
• A page that never received traffic: there the problem is upstream, in the choice of question.

Update, rewrite, merge or remove

The four routes are not interchangeable, and the choice is made before the editor is opened.

• Update when the question is the same and the page ranks: change only what has become false or incomplete.
• Rewrite when the question is the same but the page does not rank and never has. There is nothing to defend here, so the risk is low.
• Merge when two pages answer the same question. Pick the one that ranks better, move the useful content across, and redirect the retired address with a permanent redirect.
• Remove when the question no longer exists and no page on the site inherits it.

The criterion deciding between updating and rewriting is always the same: is there a ranking to defend? If yes, you work by subtraction. If no, you can start again.

What to change and what to leave alone

The body is where you can intervene most, with one warning that comes from Google: avoid changing content that is already performing well. Replace what has become inaccurate and leave the rest. What does need particular care is the page's address, the main title and the structure of the subheadings, because those are what the current ranking was built on.

If the title has to change, change it once and wait. Changing it three times in a month makes it impossible to tell which version was working.

Rules of caution

1. Do not change the address unless it is strictly necessary. If you do, the old one has to answer with a permanent redirect to the new one, direct, with no intermediate hops.
2. Change the main title once, then measure.
3. Keep the subheadings that receive citations or links. They are the points other pages and answer systems have hooked onto.
4. Isolate the variable. One substantive change at a time on a page that ranks, not a full rebuild.
5. Plan the way back. Save the previous version before publishing: if the ranking is worse at the end of the measurement window, you have to be able to return to the starting point without reconstructing it from memory.

The question of dates

It always comes up, and Google's answer is less ambiguous than people think.

On one side there are Google's guidelines on publication dates: it recommends specifying datePublished for first publication and dateModified for the last update in the structured data, in ISO 8601 format, and above all keeping the date visible to users consistent with the one in the markup. The practices Google names as wrong are specific: future dates, or the date of the event described in place of the real one, and different, contradictory dates on the same page.

On the other side, moving the date is not enough. On its page about myths and facts about crawling Google writes that content is rated by quality and not by how old it is, and that there is little value in artificially making pages look fresh by making trivial changes and bumping the date.

The practical rule that falls out is simple. The update date moves when the content has substantively changed. A date moved without real changes is a false statement about the document, and it counts as one for engines and readers alike.

How to verify the effect

An update is measured on the same phrase and the same period of the year, because seasonality explains a good share of the variation that gets attributed to edits. The right comparison is year on year, not month on month, which is why the report is worth setting to the sixteen months available.

It also takes patience with the window. Before four weeks the data is still noise, and a decisive correction on day ten destroys the only evidence you have.

How to measure it properly

1. Compare against the same period the previous year, not against last month.
2. Look at average position and clicks on the main phrase, not the page's total traffic.
3. Wait at least four weeks before intervening again.
4. Write down what you changed and when. In three months it is the only thing that will tell you what to attribute the result to.
5. If the ranking got worse and your edit is the cause, go back rather than adding another one.

Updates and AI answer engines

Here we leave the documentation and enter what we observe in the work. A badly updated page is an extra problem once answer systems are involved — not because they reward freshness in itself, but because they retrieve portions of text: if a section holds an old figure next to a new one, or a date saying one thing and content saying another, that contradiction travels inside the citation.

The practical consequence is about how you update rather than how often. Replace a figure everywhere it appears, instead of adding a paragraph that corrects the one above. Keep the main answer in a block that stands on its own after the edit too. And avoid the archaeological layers — pages where you can tell three different authors added three sections over three years without rereading the earlier ones.

Common questions