Content Audit Template and Redirect Map
Every URL gets a decision, the evidence behind it, and the mechanism that ships it. The redirect map comes out tested.
Free download · No account needed
Redirect Map · Aldermoor Software · staging test run 1 of 2
699 rules in. 693 out. 297 sources changed first.
| Rules at staging | 699 (411 merges, 288 equity retirements) |
|---|---|
| After variant expansion | 1,445 |
| Failed an assertion | 153 rules, plus 214 sources missing a variant |
The seven assertions
| # | Assertion | Result | What was actually wrong |
|---|---|---|---|
| 1 | Code is 301 or 308 | 31 failed | The CDN's bulk importer defaults every uploaded row to a temporary redirect |
| 2 | Mechanism is server side | 43 failed | The legacy resource host serves a five second meta refresh, treated as a redirect since 2019 |
| 3 | First hop lands on a 200 | 58 failed | Chained through rules left over from the 2019 migration |
| 4 | No loops | 2 failed | Two duplicate clusters resolved twice, by two people, with no survivor named |
| 5 | No target is also a source | 4 failed | Born as chains. Pointed at URLs the same map retires elsewhere |
| 6 | Target is the nearest surviving page | 19 failed | Pointed at the homepage. Thirteen had an obvious target nobody had looked for |
| 7 | Every source variant covered | 214 failed | Trailing slash form returns 404 on this server rather than resolving to the rule |
Resolution
Six of the nineteen rules under assertion 6 had no plausible target and became 410s, which is the honest mechanism for a URL with nowhere to go. The map shipped at 693 rules.
Not one of these failures is visible in a browser. All of them are visible as a traffic loss eight weeks later, in an audit that had already been declared finished.
Every content audit template on offer is an empty spreadsheet with a dropdown. Columns for the URL, the traffic and the word count, then a keep, update, consolidate or delete column somebody fills in by hand. The labels are the easy half. What happens to a page is decided by which mechanism carries the label out, and there are five of them: a permanent redirect, a temporary redirect, a 410, a noindex, and leaving it alone. They do five different things.
Google publishes what each one does. Permanent redirects show the new target in search results and temporary redirects show the source page, so a merge shipped as a 302 leaves the old URL in the index and looks identical in a browser. A 410 and a 404 are handled the same way, so the code everyone agonises over changes nothing. A 410 on a tag archive comes back on the next build. This pack decides the mechanism, then tests it.
The worked example reconciles 2,486 URLs from three sources that disagreed by 639. It retires 63 per cent of a site for nine tenths of one per cent of its traffic, and its redirect map failed 153 rules on the first staging run. Pair it with the full Search Console export past the thousand row cap and the crawl audit ranked on traffic lost. The marketing workspace keeps the inventory between quarters, so the next pass is 200 rows.
What you get
Content Inventory
One row per URL with the seven inputs a decision actually rests on, then the decision, the mechanism and a sentence of evidence naming the figures that settled it. Somebody reading only the evidence column can disagree with a specific number, which is the only kind of disagreement worth having about an audit.
Decision Method
The seven rules in the order they run, and the order carries more of the value than the rules do. Duplicate clusters resolve before any keep or update call, because a page with real traffic that is a worse version of another page is a redirect, and only cluster-first ordering ever surfaces that.
Retirement Mechanics
What a permanent redirect, a temporary redirect, a 410, a noindex and a robots.txt disallow each actually do, with Google's documentation behind every line. Including why disallow is not on the list: Google's own page on removing a page says not to use it to block a page, and a URL that is never crawled never has its noindex read.
Redirect Map
One rule per source URL, expanded to cover both slash forms and the case variants your server resolves, carrying the first status, the hop count, the final URL and the final status. A map without those four columns has not been tested, whatever anyone clicked in a browser.
Redirect Map Tests
Seven assertions run against a staging deploy before anything ships, each with the diagnosis behind the usual failure. A temporary code means an importer defaulted that way. A chain means an old migration's rules are still live. A loop means a cluster was resolved twice with no survivor named.
Retirement Queue
Everything that is not a redirect, batched by mechanism and sequenced, so the work reaches engineering as three deploys rather than sixteen hundred tickets. With the verification and the owner on each batch, and blocked batches visible instead of quietly dropped.
Overrides and Inventory Sources
Every row where a human overruled a rule, with the rule, the person and the reason. Plus where the URL list comes from and why one source is never enough: on the worked example the crawl found 565 URLs the CMS export does not list, and 74 more earning clicks appear in neither.
How it works
- 1
Send the exports
A CMS export, a crawl, twelve months of Search Console performance by page, the page indexing report, and referring domains per URL if a backlink tool is already paid for. A partial set works.
- 2
River reconciles first
Three sources, one inventory, and the URLs that appear in exactly one of them read individually. A page earning clicks that nothing links to is invisible to a crawl by definition.
- 3
Every URL gets a decision
The rules run in order, you set the thresholds, and each row carries the numbers that decided it. Then the figure that gets an audit approved: the retire set's share of the year's clicks.
- 4
The map ships tested
Merges and equity retirements become redirect rules, expanded for variants and run against the seven assertions on staging. Everything else batches by mechanism, with a dated check at fourteen and ninety days.
Frequently asked questions
Is it free, and what format are the downloaded files?
Free, and no account needed for the download. Documents arrive as .docx and every sheet as .csv, so they open in Word, Pages, Google Docs, Excel, Numbers and Sheets with no conversion step. Most people open the inventory and the redirect map first.
What does Edit with AI actually do?
It installs this pack as a private Space and River fills it from your own exports. It reconciles your URL lists, applies the rules at thresholds you set, writes the evidence on every row, builds your redirect map and tests it. The worked example is replaced by your audit.
Should I use a 410 or a 404 for pages I delete?
It changes nothing at Google. The status code reference states that every 4xx except 429 is handled identically, so both remove the URL from the index at the same speed. Use 410 where your platform makes it easy, because it is the honest code for content deliberately removed, and stop worrying about it.
Can I just redirect everything I delete to the homepage?
No, and the assertion that catches it is the reason the map gets tested. A homepage target means nobody looked for a nearest page. On the worked example nineteen rules did this: thirteen had an obvious target somebody had skipped, and the remaining six became 410s, which is the honest mechanism for a URL with nowhere to go.
How do I verify a redirect actually landed?
With a request that reports the whole status chain, not with the URL inspection panel. Google's status code documentation states that its inspection tooling does not follow redirects, so it reports on the URL you gave it and stops. That is the wrong question, and it is why redirect maps ship broken.
My site has thousands of tag and author archives. Do those get deleted?
They cannot be. The theme regenerates them on the next build, so a 410 is a URL that comes back and takes a crawl error with it. noindex is the only mechanism that applies, and rule 1 claims them before any traffic figure is read. On the worked example that was 141 rows.
Does this pick which of two pages ranking for the same query survives?
For near duplicates in your own inventory, yes, and it names the survivor before writing either rule. Where two distinct pages compete on live ranking data across a whole site, the cannibalization review does that job on query-level evidence and hands back the redirect map.
Find out what share of your traffic the deletions actually cost
Send your exports. River reconciles them, decides every URL with the evidence on the row, and hands back a redirect map that has already been tested.
Edit with AI