Open your firm’s car accident page in one browser tab. Open the car accident page for the third largest suburb you serve in another. Now read them side by side, paragraph by paragraph. If the only differences are the city name in the headline, the city name in the first sentence, and a courthouse address bolted onto the bottom, you do not have thirteen pages. You have one page and twelve photocopies, and Google is being asked to choose among them every time somebody types a query you care about.
The result is rarely a dramatic penalty. It is something quieter and more expensive. Google picks a canonical for the cluster on its own, and the page it picks is often not the one you wanted ranking. Impressions scatter across near identical URLs, none of them accumulate enough engagement or links to break out, and the firm spends two years wondering why a site with sixty pages performs like a site with six.
What Self Competition Actually Looks Like in the Data
Partners tend to hear about this problem as an abstraction. It is not abstract, and you can see it in Search Console in about twenty minutes.
Pull the last twenty eight days of performance data, filter to a single query such as the practice term plus your metro name, and switch to the Pages view. On a healthy site, one URL owns roughly eighty to ninety percent of the impressions for that query. On a duplicated site you will see something closer to four URLs splitting impressions at thirty, twenty eight, twenty four, and eighteen percent, with average positions clustered between fourteen and twenty two. That spread is the signature. Google is not confused about what the query means. It is confused about which of your pages is the answer, so it keeps swapping them, and none of them ever gets a clean run at the top of page one.
The second signature is instability. Export the same query weekly for six weeks. If the ranking URL changes three or more times across that window without any publishing activity on your side, you are watching Google re-pick a canonical. That flapping costs you more than a fixed mediocre position would, because every swap resets whatever behavioral signal the previous URL had started to build.
Measuring Real Overlap Instead of Eyeballing It
Before rewriting anything, quantify the problem. Reading pages by hand is unreliable, partly because lawyers are trained to notice small distinctions and will insist two pages are different because one mentions comparative negligence and the other does not.
Three measurement approaches, in ascending order of effort.
- Screaming Frog near duplicate detection. Enable Content in the crawl configuration, set the near duplicate similarity threshold, and crawl the site. The default threshold is ninety percent, which is far too forgiving for this problem. Drop it to sixty five percent so you catch pages that are meaningfully derivative rather than only the ones that are near carbon copies. Then exclude boilerplate by defining the content area, because your header, nav, sidebar, and footer are legitimately identical everywhere and will pollute every score if you leave them in.
- Siteliner or a similar whole site duplication scan for a fast first read on smaller sites. It is blunt, but it will tell you within ten minutes whether you are dealing with a twenty percent duplication site or a seventy percent one.
- A shingled comparison you run yourself. Strip each page down to body text, break it into overlapping five word sequences, and compute the Jaccard overlap between every pair. Fifty lines of Python does this for a two hundred page site in under a minute. The output is a matrix you can actually act on, because it gives you a number for every pair rather than a binary duplicate flag.
Set your action thresholds before you look at the results, so the numbers decide rather than your attachment to pages someone paid to write. A workable set of bands for law firm sites is this. Below forty percent overlap, the pages are genuinely distinct and need nothing. Forty to sixty five percent, they share a spine but carry real independent substance, and the fix is editorial trimming. Sixty five to eighty five percent, they are templated variants that need either a substantial rewrite or a merge. Above eighty five percent, one of them should almost certainly not exist as a separate indexable URL.
It is worth being precise about the risk here, because the phrase duplicate content carries more menace than it deserves. Google does not run a duplicate content penalty in the ordinary case. Ahrefs walks through what actually happens, which is canonical selection, filtering, and diluted signals rather than a manual action, in their guide to duplicate content. The damage is real, it is just mechanical rather than punitive. That distinction matters because it tells you the fix is editorial and architectural, not an apology to Google.
Deciding Which Page Should Own Each Query
This is the step firms skip, and skipping it is why rewrites so often fail to move anything. You cannot differentiate pages until you have decided what each one is for.
Build an ownership table. One row per target query, with these columns. The query itself. Monthly search volume from whichever tool you already pay for. The URL that currently earns the most impressions for it. The URL you want to own it. And a one line reason.
The assignment rules that hold up in practice look like this.
- The unmodified practice term belongs to the practice area page. Car accident lawyer, without a city, goes to the main practice page. Not to the page for the city your office happens to sit in, even though that page probably ranks for it today.
- Practice plus city belongs to the city page, but only where that city page can justify its existence. A city page for a suburb where you have handled four cases in six years and have no office, no reviews, and nothing local to say is not a page, it is a liability.
- Sub topics belong to sub pages, not to the city variants. Questions about rideshare collisions, uninsured motorists, or delayed injury symptoms should each own a URL of their own, linked from the practice page. Duplicating those explanations into every city variant is exactly how overlap reaches eighty percent.
- When two pages both have a legitimate claim, the older, better linked one usually wins. Check referring domains before you decide. Handing a query to a newer page and redirecting away the one with fourteen referring domains is a self inflicted wound.
Expect this exercise to shrink your page count. A firm with three practice areas and twelve city variants each has thirty six geo pages, and it is common for eight to fifteen of them to have no defensible reason to exist. That is not a failure of the audit. That is the audit working.
What Genuinely Differentiates a City Page
Once ownership is settled, the rewrite question becomes concrete. What can honestly go on a city page that cannot go on the practice page.
Swapping the city name is not differentiation. Neither is a paragraph about the city’s population and founding date, which reads as filler to a reader and as templated padding to a search engine. The material that actually distinguishes a local page falls into a small number of categories, and if a page cannot fill at least three of them, it probably should not be a standalone page.
- Jurisdictional mechanics that actually differ. Which court hears the case, filing conventions, how long the local docket runs, whether the venue has a mandatory settlement conference, how the county handles municipal liability claims. This is real, checkable, and useful, and it is different from county to county in a way population statistics are not.
- Local hazard and incident specifics. The intersections and highway segments that actually generate the collisions you handle, drawn from state DOT crash data rather than invented. A named corridor with a two sentence explanation of why it produces a particular injury pattern is worth more than four hundred words of generic negligence law.
- Local institutional detail. Which hospitals and trauma centers treat the injured, how records requests work with the dominant local health system, which police department writes the reports and how long they take to release.
- Firm specific proof tied to that place. Reviews from clients in that area, case results where the venue is identifiable and disclosable, the office address if you have one, and the travel arrangement if you do not. Say plainly how you serve a place you do not sit in. Prospects respect that far more than an implied storefront.
- Locally distinct questions. The questions people in that community actually ask, which differ more than firms expect between a dense urban core and a commuter suburb.
A city page built from that material lands somewhere between seven hundred and twelve hundred words and shares maybe thirty percent of its language with the practice page, mostly the shared legal fundamentals that genuinely belong in both places. That is a healthy number. Chasing zero overlap produces awkward writing and helps nobody.
Rewriting Without Drifting Into Doorway Territory
There is a line between a legitimate local page and a doorway page, and mass rewriting is exactly the moment firms cross it. The test is whether the page exists to serve someone in that place or to catch a query and funnel them somewhere else. If your rewrite plan is to generate substance at scale rather than to write it, you are on the wrong side of that line. We covered how to keep local pages defensible in building geo landing pages without doorway risk, and the constraints there apply to every rewrite in this project.
The related trap is deciding that the answer to twelve thin city pages is a hundred and twenty thin pages built from a spreadsheet. Templated page sets can work, but only when the template is fed by data that varies substantively per row, which is a different discipline from filling a paragraph with variables. That distinction is worked through in programmatic page sets beyond city pages.
Consolidating the Pages That Should Not Survive
For every page you decide against, the move is a 301 redirect to the page that owns its query, not deletion and not a noindex tag. Noindex keeps the page alive, keeps it collecting internal links, and throws away whatever authority it accumulated. Redirect it.
Before redirecting, harvest. Read the page for the two or three genuinely local details it contains and move them into the surviving page. Most thin city pages have one real sentence in them, and it is worth keeping.
Then fix the internal links. This is the step that gets dropped and it matters more than the redirect itself. Crawl the site again, find every internal link pointing at a redirected URL, and repoint it at the destination. Leaving twenty internal links chained through redirects tells Google you were not confident enough about the consolidation to update your own navigation.
If your firm runs several practice areas, do the same ownership work across practices, not only across cities. The overlap between a general personal injury page and a car accident page is frequently worse than the overlap between city variants, and the fix follows the same logic laid out in running multi practice law firm SEO without dilution.
Measuring Whether the Rewrite Worked
Set expectations before the work ships, because the early signal looks like a problem. Page count drops, total indexed URLs drop, and total impressions often dip for three to five weeks as the redirected pages fall out and the surviving pages absorb their signals. If your partner meeting dashboard tracks indexed pages as a success metric, change it now.
What to actually watch, at thirty, sixty, and ninety days.
- Impression concentration per query. Re-run the Search Console query to page breakdown. The number you want is a single URL holding above eighty percent of impressions for its assigned query. That consolidation usually shows up before position improves.
- Ranking URL stability. Zero or one URL changes per query across six weeks, versus the three or more you measured going in.
- Average position on the assigned query for the assigned page only. Not sitewide average position, which is close to meaningless after a consolidation because the URL mix changed.
- Consultations, filtered by landing page. The only number the partners actually own. Twelve pages splitting six inquiries is worse than four pages producing nine.
The Honest Tradeoffs
This work is slow and it is manual. A serious rewrite of a city page takes two to four hours including the jurisdictional research, and there is no version of this that a content tool completes for you at volume, because the entire value of the page is the specificity a general purpose tool does not have.
Consolidation also costs coverage. Redirect eight city pages and you will lose some long tail impressions from those towns, permanently. That is usually a good trade, because those impressions were converting at close to nothing, but it is a real loss and you should decide on it with open eyes rather than discover it in month two.
And the timeline is quarters, not weeks. Google needs to recrawl, process the redirects, and re-evaluate the surviving pages. Ninety days is a fair first read, six months is a fair verdict. Anyone promising you rankings on a shorter clock is selling something other than this.
Stop Competing With Yourself
The fastest way to find out where you stand is to measure. Crawl the site with near duplicate detection at a sixty five percent threshold, pull the query to page breakdown for your five most valuable terms, and see how many URLs are splitting each one. If four pages are dividing the impressions for the query that pays your firm’s bills, you already know what the next quarter of content work is. Rubiks runs this overlap and ownership audit as the first step of law firm content engagements, and the deliverable is the ownership table plus a page by page keep, rewrite, or redirect call. Talk to Rubiks and we will tell you exactly how many of your pages are competing with each other before you commission a single rewrite.