Content Audit Before a Website Redesign: Keep, Fix, Merge or Remove Each Page Using 6 Criteria and a Template

In a redesign, design and CMS discussions usually come first. A few meetings in, someone notices nobody has decided how much of the current site moves to the new one: ten years of news, a stalled blog, pages nobody can explain since their author left. "Move everything" and "drop everything" both lose you something.

Making that call page by page, with reasons, is a content audit: a content inventory (the list) plus an evaluation. Many agencies treat it as step one of a redesign.

Following Google Search Central's guidance, this article covers collecting every URL, six criteria for judging pages, deciding to keep, fix, merge or remove, using 410, 301 or noindex, multilingual sites and effort by site size. Redirect setup for changed URLs is covered in "Website migration SEO", and the overall redesign schedule in "The website redesign process and timeline".

1. What a content audit is and why to do it before a redesign

The inventory and the audit are separate jobs

Nielsen Norman Group defines a content inventory as a list of all the digital content you currently have, and a content audit as the evaluation of the quality of the content on that list. The first can be automated; the second needs a person to read.

The same article lists inventory fields such as title, URL, owner, topic, format and last-modified date. The audit template in section 9 adds metrics and a decision column to those fields.

What happens when you skip it

Skipping the audit causes three problems. First, everything is migrated and the new site is cluttered from day one. Second, everything is dropped and the entry points people used from search and external links disappear. Third, content deadlines arrive before anyone has decided which pages are moving.

Before it is an SEO task, the audit sets the content volume and who writes it.

When to do it

Audits are not only for redesigns, but before one they pay off most, because they shape URL structure and migration scope.

Redesigns and CMS switches cover every URL; smaller triggers need only the affected section.

2. Does deleting old content really improve rankings?

What Google's Search Liaison said

The question we hear most is "Will deleting old posts help SEO?" The idea spread partly through US tech publisher CNET, which, Gizmodo reported in August 2023, had deleted thousands of old articles to perform better in search.

Google's Search Liaison, Danny Sullivan, responded on X and rejected the idea that Google dislikes old content. As quoted by Practical Ecommerce, he wrote "That's not a thing!", adding that Google's guidance does not encourage this and that older content can still be helpful.

What the official guidance says

Google's page on helpful, people-first content lists self-assessment questions. Two apply directly.

One asks whether you are changing the dates of pages to make them seem fresh when the content has not substantially changed. The other asks whether you are adding a lot of new content or removing a lot of older content mainly because you think it will help rankings.

Google also said its March 2024 core update was expected to cut low-quality, unoriginal content in search results by 40%, and later added that the rollout completed on 19 April with a 45% reduction. The goal is not removing "old" pages but fewer unhelpful pages and stronger helpful ones.

Using the self-assessment questions in your audit

The questions work as criteria for the accuracy and freshness columns.

The test: can you state each decision's reason in one sentence from the user's view? Hence the "reason" column.

When removal does make a difference

Removal does matter in some cases. Google's guide to managing crawl budget is aimed at sites with over a million pages that change about weekly, or sites with over 10,000 pages that change daily. For those sites it recommends returning 404 or 410 for permanently removed pages, eliminating soft 404s and consolidating duplicate content.

For a company site of a few hundred pages, crawl efficiency is rarely an issue. The real gains are that users stop landing on outdated information and there is less to update, translate and maintain.

3. Finding every URL: four sources

XML sitemap: the pages you meant to publish

Start with the XML sitemap. Look for /sitemap.xml or the Sitemap: line in robots.txt. According to Google, a single sitemap can hold up to 50,000 URLs or 50MB uncompressed; larger sets are grouped with a sitemap index. However, hand-placed HTML files, old PDFs and campaign pages are often missing.

Crawler: the pages you can reach through links

Next, use a crawler that follows internal links. Screaming Frog SEO Spider, a common choice, is free for up to 500 URLs; the paid licence costs £199 a year (checked 28 October 2026). It exports status codes, titles, canonicals and noindex in one pass. It cannot find orphan pages with no internal links.

Search Console: pages Google knows about, plus backlinks

The Search Console performance report shows pages that received impressions and clicks. Tables show up to 1,000 rows, though. For more, use the Search Analytics API: Google says you can page through results 25,000 rows per request, with up to 50,000 rows per day per search type.

Search Console's Links report shows which of your pages have the most links from other sites. The table is limited to 1,000 rows, but the export button gives you up to 100,000 rows. Pages with backlinks change the removal decision, so always export this data.

GA4: pages people actually visit, and enquiries

GA4's landing page report shows where people arrive from every channel, not just search (social, email, ads, direct), and which key events followed, such as enquiry form submissions.

Watch the retention setting. GA4 keeps event data for 2 or 14 months (up to 50 months in GA4 360). Google says this setting only affects explorations and funnel reports, not standard aggregated reports. Check it is not still set to 2 months.

Comparing the four sources

The table compares the four.

Any single source misses pages, especially old pages with backlinks but no search traffic.

Merging them into one master list

Combine the four exports in one spreadsheet, normalise the URLs (trailing slashes, http versus https, www or not, tracking parameters), then remove duplicates. Which sources a URL appears in tells you something about the page. A page in neither Search Console nor GA4 is a page nobody visits. A URL only in GA4 may be an ad or email landing page; check with its owner first.

4. The six criteria

Criterion 1: traffic (search and referral)

The first criterion is whether people visit the page. Look at Search Console clicks and GA4 landing page sessions over the last 12 months. Some pages are seasonal (holiday opening hours, year-end services), so do not decide on three months of data.

Criterion 2: backlinks

The second is whether other sites link to the page. Deleting it leaves those visitors at a dead end and loses a search signal. If there are backlinks, redirect the page with a 301 to the closest related page even if you remove it. Links from clients, municipalities or the press count heavily even if few.

Criterion 3: contribution to enquiries

The third is whether the page leads to enquiries. Check GA4 key events by landing page and by pages viewed before an enquiry.

Pages read just before people get in touch (pricing, case studies, FAQs) stay even if traffic is low. Judging on search numbers alone can remove the pages that close the deal.

Criterion 4: accuracy

From here on, a person has to read the page. Check whether prices, services, regulations, legal information, contact details, addresses and names are still correct. Read pages with factual content, such as prices and rules, first to deal with the riskiest pages quickly.

Criterion 5: freshness

The fifth is whether the content fits the current situation. Accuracy asks "is anything wrong?"; freshness asks "does it answer today's reader's question?" A 2019 article on "how to start remote working" may contain no errors yet miss what readers care about now.

Note that the publication date alone is not a reason to act. As section 2 showed, Google does not recommend deleting content for being old. A ten-year-old page about how the company got its name can still be useful.

Criterion 6: duplication

The sixth is whether several pages cover the same topic: a "price change notice" rewritten every year, three blog posts answering the same question, the same PDF at separate Japanese and English URLs. Duplicates split signals in search, and readers cannot tell which one is current.

Google's guidance on consolidating duplicate URLs gives reasons for a canonical URL, including consolidating signals and saving crawl time. It describes redirects and rel="canonical" as strong signals and sitemap inclusion as a weak signal. When the audit finds duplicates, choose the page to keep and redirect the rest to it with a 301.

5. Keep, fix, merge or remove: the decision rules

The order of the questions

A fixed order keeps decisions consistent across people. We recommend duplication, then evidence of value, then accuracy.

Duplication comes first because deciding where pages merge before anything else reduces rewriting.

Even if the second question finds no evidence of value, pages the business or the law requires, such as the company profile, privacy policy and accessibility statement, are never removed. Mark them as "keep" before the audit and check only their accuracy.

The four outcomes

Every page gets one of four outcomes.

"Fix" and "merge" often happen together. If three old articles become one new article, record the target as "fix" and the other two as "merge".

Pages that are hard to classify

Most time goes on pages that do not fit neatly.

When in doubt, ask: what should someone who lands on this URL today see? If the answer is "this content", keep or fix it. If it is "another page", use a 301. If it is "nothing", remove it.

6. When to use 410, 404, 301 and noindex

How Google treats 404 and 410

Google's documentation on HTTP status codes says that all 4xx errors except 429 are treated the same: Google's crawlers tell the next processing system that the content does not exist, and in Search a previously indexed URL is removed from the index. 410 (Gone) gets no special treatment among the 4xx codes.

The HTTP specification (RFC 9110), meanwhile, defines 410 as meaning the resource is no longer available and the condition is likely to be permanent. A practical rule: use 410 when you want to say the removal is permanent, and 404 when your CMS makes 410 difficult.

Google's guide to site moves with URL changes also says that if you are not moving all your old content, make sure those URLs correctly return a 404 or 410.

Merge with a 301, and do not send everything to the homepage

When moving a page's role to a closely related page, use a permanent redirect (301 or 308). Google says a server-side redirect has the highest chance of being interpreted correctly.

Avoid sending every page without a destination to the homepage. Google's site move guide says not to redirect many old URLs to one irrelevant destination, such as the new site's homepage, and that this might be treated as a soft 404. The same guide says to keep redirects "for as long as possible, generally at least 1 year".

noindex: keep the page, hide it from search

noindex keeps a page published but removes it from search results. Google explains that when it is set with a <meta> tag or HTTP header, the page is dropped from results even if other sites link to it.

A noindex rule on a page blocked by robots.txt is never read, and Google says recrawling can take months. noindex suits pages that users need but do not need to find through search.

The Removals tool is temporary

Search Console's Removals tool hides a page on your site from Google results within about a day. But Google says these requests last about 6 months. A permanent fix needs the page removed or updated, password-protected or set to noindex. Use the tool only for urgent cases, such as a page exposing personal data.

Comparing the options

Here is what each option does.

Choose by what visitors should see, and give your error page a search box and key links.

7. Multilingual sites: when only one version is out of date

A common situation

On multilingual sites, one language is often updated while the other is left behind. If the two are connected with hreflang (the markup that tells search engines which pages are language versions of each other), you are also declaring two pages with different content to be versions of the same page.

hreflang only works both ways

Google's guidance on localized versions says each language version must list itself as well as all other language versions, and that if two pages do not both point to each other, the tags will be ignored. If you remove or redirect one version, remove it from the other version's hreflang too.

Comparing the options

The table sets out options by situation.

When prices differ by market, do not force an hreflang link; independent pages mislead nobody.

Before filling gaps with machine translation

When one side is missing pages, it is tempting to fill the gap with machine translation. Google's spam policies list publishing other sites' content through automated transformations such as translating as an example of scaled content abuse. Your own pages are a different case, but mass unreviewed machine translation still adds low-quality pages.

Choose pages to translate from the "keep" and "fix" rows of your audit sheet, in order of traffic and enquiries, and publish them after a person has reviewed them.

8. Pages to review for personal data and legal reasons

GDPR and storage limitation in Europe

In Europe, the audit doubles as a personal data review. Article 5 of the GDPR (the EU General Data Protection Regulation) sets out principles including data minimisation (only what is necessary), accuracy (kept up to date where necessary) and storage limitation (not kept in identifiable form for longer than necessary).

Photos of former staff, photos of event attendees and unused application forms should be judged by whether there is still a basis for publishing them, regardless of traffic.

Pages where accuracy has legal consequences

Errors in prices, contract terms or certifications can cause problems with consumers. Recording which legal requirement a page exists to meet makes it easier to prioritise checks. If the European Accessibility Act applies, old PDFs need reviewing too (see "The European Accessibility Act and your redesign").

9. Content audit template

Column design

Use one spreadsheet. Start with these 13 columns and adjust.

The "reason" column is the only record that answers "why did we delete this?" six months later.

Example rows

Here is an example for a fictional site. The numbers are for illustration only.

Required pages stay despite few clicks; the old plan has backlinks, so it redirects to its successor.

Feeding the sheet into the new site's design

Once the sheet is done, assign the "keep" and "fix" pages to the new site's structure. That becomes the new site's page list and content plan, and the "merge" and "remove" rows become the redirect map.

10. Effort and cost by site size

Assumptions

There is no published market rate for content audits alone. The figures below are practical estimates from project work, not statistics. We assume about 3 minutes per page for decisions that can be made from the data and about 10 minutes for pages that need reading. Rewriting is not included.

Costs use Dutch hourly rate surveys (freelance web developers €73, average agency base rate €106, both excluding BTW, the Dutch VAT), converted at the European Central Bank reference rate of 28 October 2026 (€1 = ¥175.86). The rate sources are in "Website redesign cost".

Estimates by size

The table combines inventory and measurement, decisions and setting up redirects and status codes.

Above 1,000 pages, judge tag and auto-generated pages by rule and read only top pages and pages with factual content.

Quote rewriting separately

Rewriting pages marked "fix" is a separate cost from the audit. Allow 1 to 4 hours per page, or €73 to €424 (about ¥13k to ¥75k) at the rates above. Since volume is unknown until decisions are made, commission the audit at a fixed price first and quote rewriting once you know the count.

11. Common mistakes

Mistakes and how to prevent them

The table lists common audit mistakes, what happens and how to avoid them. The first three in particular happen when Google's guidance from section 6 is overlooked.

12. Worked examples (fictional)

The following three examples are fictional and do not describe real companies.

Case A: a Japanese professional services firm, 80-page site

Ten articles explain rules from before a legal change. The four still getting traffic were marked "fix" and the other six "merge" into a new explainer. Thirty news posts were grouped into three yearly archives, each old URL redirected there with a 301. Audit and setup: about 20 hours, plus about 16 hours of rewriting.

Case B: a Japanese-owned service company in the Netherlands, 400 pages in two languages

Half of the English pages show prices from three years ago. English gets a third of the Japanese traffic but 40% of enquiries. The top 30 English pages for enquiries were marked "fix". Fifteen English pages for services not offered in the Netherlands were removed and taken out of the Japanese pages' hreflang. Audit and setup: about 50 hours, plus about 90 hours of rewriting and translation review.

Case C: an online shop selling into the EU, 3,000 URLs

Of 1,200 products, 400 are discontinued and about 60 of those have backlinks; there are also 1,500 tag listings and 300 blog posts. Tag listings were judged by rule (e.g. "fewer than three products"); people read only the blog and the 60 linked products. About 250 discontinued products were redirected to successors with a 301; about 150 with neither a successor nor backlinks return 410. The work took about 120 hours.

The larger the site, the more the time depends on how tightly you limit what people read.

13. Content audit checklist

Finally, here is what to confirm at each stage.

Collect

• Exported URLs from the XML sitemap, a crawler, Search Console and GA4
• Normalised URLs and removed duplicates
• Exported backlink data from Search Console

Measure

• Added clicks, sessions and key events for the last 12 months
• Flagged required pages (company profile, terms, legally required information)

Decide

• Judged pages in order: duplication, evidence of value, accuracy
• Wrote a one-sentence reason from the user's view for every row
• Agreed with the business owner how to handle discontinued services and old prices
• Judged language pairs together, in both directions

Act

• Redirected merged pages with a 301 to the closest related page
• Removed pages return 410 or 404, with no bulk redirect to the homepage
• noindex pages are not blocked in robots.txt
• Removed language versions are taken out of the other side's hreflang
• Set the next review date and owner

Summary

• A content audit has two stages: listing every URL and evaluating each page
• Build the list from the sitemap, a crawler, Search Console and GA4 together, not from one source
• Judge pages on traffic, backlinks, enquiries, accuracy, freshness and duplication, in the order duplication, evidence of value, accuracy
• Google does not recommend deleting content because it is old. The reason to remove a page is that it does not help users
• Use 301 to merge, 410 or 404 to remove and noindex to keep a page out of search. Avoid bulk redirects to the homepage and combining noindex with a robots.txt block
• On multilingual sites, when you remove one version, remove it from the other version's hreflang too
• Expect 8 to 15 hours for up to 50 pages. Quote rewriting separately once the decisions are made

Sources

Google's official guidance

• Creating helpful, reliable, people-first content — Google Search Central: self-assessment questions
• How HTTP status codes, and network and DNS errors affect Google Search — Google Search Central: handling of 4xx codes
• Redirects and Google Search — Google Search Central: permanent redirects
• Site moves with URL changes — Google Search Central: bulk redirects to the homepage and soft 404s
• Block Search indexing with noindex — Google Search Central: how noindex works and the robots.txt caveat
• Remove information from Google — Google Search Central: Removals tool lasts about 6 months
• How to specify a canonical URL — Google Search Central: reasons to consolidate and signal strength
• Large site owner's guide to managing your crawl budget — Google Search Central: which sites it applies to
• Tell Google about localized versions of your page — Google Search Central: hreflang return links
• Spam policies for Google web search — Google Search Central: scaled content abuse
• Build and submit a sitemap — Google Search Central: 50,000 URLs / 50MB
• New ways we're tackling spammy, low-quality content on Search — Google: March 2024 core update

Data collection

• About Search Console data — Search Console Help: 1,000-row table limit
• Links report — Search Console Help: most-linked pages
• Getting all your data — Search Console API: 25,000 rows per request
• Data retention — Google Analytics Help: 2 or 14 months
• SEO Spider pricing — Screaming Frog: free version 500 URLs

Audit approach and reporting

• Content Audits — Nielsen Norman Group: inventory versus audit (27 September 2020)
• SEO and Old Content — Practical Ecommerce: quotes from Google's Search Liaison (17 August 2023)
• Google warns against content pruning as CNET deletes thousands of pages — Search Engine Land: the CNET deletions and Google's response
• Cool URIs don't change — W3C: why URLs should not change or disappear
• RFC 9110: HTTP Semantics, 410 Gone — IETF: definition of 410

Law, rates and exchange rates

• Regulation (EU) 2016/679 (GDPR) — EUR-Lex: Article 5 principles
• Wat verdient een web developer (zzp)? — Knab: freelance web developers €73
• Benchmark report Agencies Q4 2025 — Simplicate: average agency base rate €106
• Euro foreign exchange reference rates: Japanese yen — European Central Bank: reference rate on 28 October 2026, ¥175.86

All sources were checked on 28 October 2026. Google's guidance and tool limits and prices can change, so check the official pages before you start.