Content conflicts: your pages that overlap
A conflict is a pair of pages on your website that compete for the same Google query: they are so similar, or so clearly target the same thing, that Google doesn’t know which one to show and both do worse. This is cannibalisation. The ERP looks for them by itself each time it reads your website and leaves them in a tray so that you can decide what to do with each one. It is in Marketing → Search visibility, Conflicts tab.
And better than fixing a conflict is not creating it: each piece in the editorial plan is compared with what is published and with what is planned before it is written. See and a piece that isn’t on the website yet?.
How does the ERP know that two pages overlap?
In three ways:
- Similar titles. It compares the title of each page (the post’s and the one Google sees) with that of all the others. The words that say what the title is about count: ‘of’, ‘how’ or ‘everything you need to know’ don’t count, and a word weighs more the fewer pages on your website use it. The brand, which appears in almost every title, weighs almost nothing. What comes before the colon or the dash (‘Time tracking when working from home: guidelines and mistakes’) weighs twice as much as what comes after: it is the query; what comes after is the hook.
- The same target query. Two pages that have been given the same target query on the map (ignoring capitals and accents) always overlap.
- A query shared in Google. If the website is connected with Google, the ERP looks at the real Search Console queries: two of your pages that Google shows for the same query and that share its impressions conflict. They share them when the second gets at least 20 % of the impressions of the first (and a minimum of 10 in 28 days); below that, the first dominates and Google has already chosen. It is the way that measures what really happens, even if the titles aren’t similar at all.
If the ERP has the assistant’s meaning model installed, it also finds titles that say the same thing in other words: ‘Free Verifactu: can you get it?’ and ‘Is it possible to download Verifactu?’ hardly share any words, but they ask the same thing.
Which pages can conflict?
Only those Google can show:
- Published. A draft, a private page or a withdrawn one doesn’t compete with anyone.
- Indexable. A page with ‘noindex’, or whose canonical points to another address, isn’t in Google: it can’t cannibalise.
- In the same language. The English version of a page doesn’t compete with the Spanish one, and the translations of the same page never conflict with each other.
And the blog categories (they are navigation, not destinations) and what is in Other (the legal notice, the contact page, what the rules couldn’t classify) don’t compete by title. By target query and by query shared in Google they do: if you give a category a query, it means you want it to target it, and if Google shows it instead of your pillar page, it is taking visits away from it.
How do I read a conflict?
The tray shows, for each conflict:
- The two pages, with their scope.
- Why: ‘Similar titles’, ‘Same target query’ or ‘Query shared in Google’, and the evidence: which words they share (‘They share ‘time’, ‘tracking’ and ‘homeworking’ · similarity 0,65’), that they say the same thing in other words, which query both target, or which Google queries they share, with the impressions and position of each page (‘time tracking’: A 1.200 impressions, position 7,3, and B 640, position 11,4).
- The score, from 0 to 1: how much they overlap. A title conflict starts at 0,45; 1 is the same title. Two pages with the same target query score 1. In a shared query, the score says how much they share it: 1 is half and half, and it never goes below 0,20.
- The status: Open (to deal with), Resolved (with its outcome, or ‘they no longer compete’ if the website fixed it) or Dismissed.
At the top, the buttons count the conflicts in each status and filter by it. The search box searches the titles and addresses of the two pages, ignoring capitals and accents, and the drop-down filters by conflict type.
Clicking a conflict opens its record with the two pages side by side: title, SEO title, scope, cluster, target query, intent, status and address of each one. It is what you need to decide which one keeps the query.
What do I do with a conflict?
You choose one of the four outcomes, which are the ways of ending a cannibalisation:
| Outcome | What is done on the website | When |
|---|---|---|
| Differentiate | Each page is rewritten towards its own query: different title, headings and approach | Both make sense separately, but they are written for the same thing |
| Merge | One absorbs the other and the other redirects to it | They say almost the same thing: two blog posts on the same topic |
| De-optimise | One stops targeting the query: it is removed from the title and headings and links to the other | One is the good one and the other has to keep existing for another reason |
| Noindex | One leaves Google’s index (‘noindex’): it stays on the website for whoever browses | A page that has no reason to appear in Google: a listing, a pricing page for another product |
Marking the outcome in the ERP doesn’t change your website: it records what was decided. The change is made in WordPress (or on your website), and the next time the ERP reads it, it will check whether it was enough.
In addition:
- Discard: it isn’t a conflict (the two pages answer different intents even though they look alike). The conflict stays in the tray as dismissed and doesn’t reopen while the two pages say the same as now.
- Open task: creates an ERP task to deal with the conflict, with the two pages and the evidence in the description. The task stays linked to the conflict and the tray shows its number.
- Reopen: if it was dismissed or resolved by mistake.
To decide you need the Changes permission on Search visibility; with View you can see the tray, but not change it.
Can a conflict come back?
Yes, and that is how the ERP checks the fixes:
- A resolved one reopens if the website changes either of the two pages and they still overlap: the fix wasn’t enough. While the website doesn’t change, it stays resolved (the fix is still to be made on the website).
- A dismissed one only comes back if either of the two pages changes (its title, its SEO title, its H1, its robots…) and they are still similar.
- An open one that is no longer detected —one of the two pages leaves the index, is withdrawn or changes its title, or Google stops sharing the query between the two— is considered resolved by itself: it appears as ‘Resolved · they no longer compete’. If they overlap again, it reopens.
A conflict that is right on the limit doesn’t close and open with each new page you publish: to close by itself, its score has to drop clearly.
When are conflicts looked for?
Each time the ERP reads your website (see sync your website), after bringing the map up to date. And whenever you want, with Detect now in the Conflicts tab: it looks for them again with the map as it is, without reading the website. Use it, for example, after giving two pages the same target query.
When a new conflict appears (or one that was closed comes back), the ERP warns with the event ‘Two pages on the company website overlap’, to which a task schedule can be attached: create a task for marketing, send a notification… The first time a website’s conflicts are looked for, you aren’t warned about each one: they were already on your website, and they are all in the tray.
And a piece that isn’t on the website yet?
Each piece in the editorial plan is compared with what is published (the published and indexable pages in the same language), with what is planned (the other pieces in the plan that aren’t discarded, including the guest posts on your website) and with the guest posts you publish on other websites, by the query they target: an article of yours on a domain with more authority can take your piece’s position. Six signals are checked:
| Signal | What it looks at | Colour |
|---|---|---|
| Same search query | Its query is already the main query of another page, of another piece or of an article of yours on another website, or another one covers it in passing | Red if it is the one the piece targets; amber if the piece or the other only covers it in passing |
| Already ranks in Google | Google already shows a page of yours for that query (Search Console) | Red up to position 30; amber further down |
| Similar titles | The words in the titles, just as between pages | Amber; red if they say exactly the same words |
| Same meaning | Titles that say the same thing in other words | Amber |
| Shared Google results | Its query and that of another owner share at least 4 of the top 10 results: for Google they are the same | Red |
| Same intent in the cluster | Another piece or page in the same cluster with the same intent | Amber |
The piece’s colour is that of its strongest signal (with no signals, green), and with it comes what to do:
- Create: nothing conflicts, go ahead.
- Improve the existing one: a page of yours already targets the same thing, or Google already shows it for that query (‘You already have /blog/verifactu/ in position 14 for ‘mandatory verifactu’: improve that page instead of creating another’).
- Merge: another piece in the plan targets the same thing: merge them.
- Differentiate: it resembles something without fully overlapping it, or it targets the same as an article of yours on another website (which can’t be improved or merged from here): give it another angle or another intent.
Each signal comes with its proof (which query, which words, how many impressions, in what position, how many results in common) and a link to see what it conflicts with. If something couldn’t be checked —the website isn’t connected with Google, the query has no saved Google results, there is no meaning model—, the record says so: a green without that data isn’t the same as a green with it.
When is it checked? When you add the piece, when you change it (its title, its queries, its intent, its cluster), when you create it with Create the missing pieces, with Check conflicts in the editorial plan, with Detect now in this tab and every night if you turn on the example schedule ‘Comprobar los choques del plan editorial cada noche’ (it starts paused, like the others: see the publishing board). A page that is published or a query that Google starts showing you for changes the colour the next day.
Accept the conflict. If marketing decides the piece is written anyway (‘the page sells and the piece teaches’), on the piece’s record type the reason and click Accept the conflict: the column changes to Accepted and doesn’t warn again as long as the piece, or what it conflicts with and why, doesn’t change. If it changes (another title, another query, a new page that also conflicts), it warns again. Warn again withdraws the acceptance. The figures in the proof (a page gaining a few impressions) don’t count as a change.
Below the tray of conflicts between pages, this tab shows the pieces in the plan that conflict and haven’t been accepted, red ones first. They are dealt with on their editorial plan record.
The thresholds (the similarity of titles and of meaning, the Google results in common, the minimum impressions and the position that makes a conflict red) are changed by marketing in Guidelines → Weights and thresholds, Conflicts when proposing a piece group.
Can I ask the assistant to do it?
Yes. The assistant looks things up in the tray (‘which pages on the website overlap?’, ‘is there cannibalisation on the blog?’, ‘what conflicts does the time tracking page have?’) and can record what was decided (‘conflict 12 is resolved by merging the two pages’, ‘dismiss conflict 7’). Like every assistant action, it shows it to you before doing it, and the result is the same as on screen.
Before proposing a piece it can also check it: ‘if I write an article about what verifactu is, does it conflict with anything on the website?’, with a title, a query or a piece from the plan. It only looks: it doesn’t create the piece or save anything.
What doesn’t the detector see?
- Two pages that answer the same query with titles that aren’t similar, if the website isn’t connected with Google. The detector compares titles; if two articles cover the same thing with very different titles and the meaning model doesn’t manage to see it, only the query shared in Google or the same target query on the map finds them.
- What the pages say inside. Titles are compared, not the text of the pages.
- A single word in common. Two titles that only share one word (‘Stock’ and ‘Maximum stock’, ‘QR in Verifactu’ and ‘QR code on e-invoices’) aren’t a title conflict, even if they are very similar: that way the tray isn’t filled with pairs that don’t overlap. If such a pair does overlap, it shows up through the query shared in Google or through the same target query.
- Queries with very few impressions. A query in which the second page doesn’t reach 10 impressions in 28 days doesn’t make a conflict: it is noise.
- In pieces, what doesn’t exist yet. A piece is compared by its working title, its queries, its intent and its cluster: what the text says when it is written isn’t seen until it is on the website. Guest posts on other websites only count as owners of the query they target (their title isn’t on your website); the rejected or withdrawn ones no longer count.
Back to search visibility.
More about Search positioning
- Guest posts
- Proposed clusters
- Connect your website with Google: Search Console and Analytics
- Query real search data: what it costs
- Cross-checks between SEO and ads
- Add your website and test the connection
- Search data: volume, difficulty and Google results
- Marketing guidelines: what marketing tells the plan
- The editorial calendar
- A piece’s brief
- The content map: what each page of your website targets
- The Looker Studio dashboard: your content map alongside what Google measures