Nobody deletes anything: the stale page problem lands on your team

Every dealer website platform is built for one direction: adding. Add a model page, add a service page, add a campaign landing page, add a specials module, add a location. Nothing in the commercial model rewards the opposite, and the opposite is what keeps a site healthy at scale.

Ask for a page to be removed and the request changes character. It becomes a production ticket, a statement of work line, or a conversation with a rep about what your tier includes. That is not sabotage. It is a service model designed around growth, staffed for builds, and priced around deliverables you can see. Which means the removal problem is not a vendor failure. It is an unassigned job that quietly becomes your department's.

How a dealer site reaches 20,000 URLs without anyone deciding to

No single decision creates the archive. Accretion does.

Run a site search against a five-rooftop group that has been on the same platform for eight years and the number is usually far larger than anyone internally guessed. That gap between the page count you believe you have and the URL count a crawler finds is the tell.

Page count has been sold as progress

Part of why this persists is that the industry measured websites by volume for a decade. Page-count production reports, "we added 40 pages this quarter," SEO retainers that invoice by deliverable volume. Vendor evaluation guides now flag exactly this as a red flag: a provider that sells by page count rather than by outcomes is selling activity, not results.

That framing survives because it is easy to verify and easy to present. Nobody wants to present a slide that says the quarter's work was nineteen removals, six redirects and a decommissioned promotions archive. Yet that is frequently the higher-value quarter.

Removal is not in the default scope

This is the part worth reading twice, because it changes budget and headcount rather than tactics.

On managed platforms of the Dealer.com class, day-to-day changes route through a production queue. Operators have complained for years that even trivial edits, down to a logo swap, require a support ticket, and that requests outside the agreed scope come back as "that is extra." The detail is old; the structure is not. Basic tiers get a self-service design editor and a ticket queue. Higher tiers add management, more modules, more reporting, dedicated contacts. What they rarely add is a content lifecycle policy: an automated job that retires expired campaigns, purges sold inventory URLs on a rule, or flags pages nobody has touched in three years.

So the honest framing of a platform contract is not "what does it do" but "what does it refuse to do, and how much does the refusal cost per request." Five questions to put in writing before renewal:

1. Does the contracted scope include page removal and decommissioning, at our tier, in writing?
2. Is there an automated retirement job for expired promotions and sold inventory URLs? What triggers it?
3. What is the turnaround on a removal request, and is removal batched or per-ticket?
4. Who owns redirect management when a URL is retired, and is the redirect map exportable?
5. Can we pull a full URL inventory on demand, including pages not linked from navigation?

A "no" to any of those is not a dealbreaker. It is a staffing input, and it needs to appear in a job description rather than in the hope that someone remembers.

What the archive actually costs

Crawl budget. Google is explicit that crawl budget is finite and that pages you should not have are competing with pages you need indexed. Permanently removed pages should return 404 or 410 rather than being blocked, because blocked URLs stay in the crawl queue far longer. Soft 404s keep being crawled. This is not theory in a category where inventory turns weekly.

Redirect chains. Years of unmanaged retirements produce chains four and five hops deep. Each hop is a fraction of link equity and a fraction of a second, multiplied across thousands of URLs.

Stale promises. A landing page from a 2021 promotion still quoting a trade-in bonus, a service page with last year's pricing, an hours page that has not matched the store since the remodel. This is the category of stale content with legal and reputational weight, and it is the one most likely to be discovered by a customer rather than by you.

Duplicate and cannibalising pages. Multiple model-year pages competing for the same query, with the newest one losing to a page from three years ago that has accumulated links.

Answer engines. Assistants do not politely prefer your current page. They read what is indexed and citable. A stale service page with a confident price is exactly the kind of source a generative answer will quote, and you will not know until a customer arrives quoting it back at you.

Analytics you cannot trust. Landing-page reports diluted across dozens of near-identical URLs, so nobody can tell which page actually earns.

Why most groups are not staffed for it

The work is not a marketing tactic. It is web operations: an inventory, an owner per URL class, a lifecycle rule, a redirect map, a recurring audit, a ticket log, and a monthly check on indexed versus intended page count. That is a named function, and in most groups it does not exist.

Headcount benchmarks make the problem concrete. Organizations in the $10M to $50M revenue band run median marketing teams of a handful of people, often one to six, wearing demand generation, content, ops and reporting at once. Dealership-level content functions are routinely a single person; one operator describes running a 55-person store with a one-person content team. Meanwhile Cox Automotive's dealership staffing research found 67% annual turnover on sales teams, with an average $10,000 cost per hire. Tribal knowledge does not survive that rate.

Which is the crux. Every page in the archive was created by someone with a reason. When that person leaves, the reason leaves, and the page becomes unfalsifiable. Nobody can prove it is safe to delete, so nobody deletes it. The rational individual decision produces an irrational collective result, one page at a time.

Agencies do not default to filling the gap either. Cleanup is unglamorous, hard to attribute and easy to argue about, so it becomes either an add-on nobody buys or a favor someone performs until they are busy.

The fix is rules, not cleanups

One-off cleanups fail because the site keeps generating. What holds is a lifecycle rule per URL class, agreed once, executed repeatedly.

The single highest-leverage habit is the cheapest one: no campaign page ships without an expiry date and a successor URL. It converts a future archaeology project into a scheduled task, and it costs one field in a request form.

A 30-day starting sequence

1. Export every URL: sitemap, a full crawl, and the Search Console page inventory. Reconcile the three.
2. Classify into the classes above. The count per class is the argument you will need for budget.
3. Fix the high-risk stale pages first: pricing, promotions, hours, financing claims. That is risk reduction, not SEO.
4. Ticket the removals in one batch, log the turnaround, and use the real SLA in next year's renewal conversation.
5. Assign one named owner with a defined number of hours per month. Not "marketing." A person.
6. Report three numbers monthly: indexed URLs against intended URLs, 404 volume, and average redirect chain length.

The uncomfortable conclusion

Platforms in this class will keep building. That is what they are for, and in fairness, a dealer group with forty rooftops genuinely needs a production engine. But a production engine with no retirement process is an archive with a homepage attached, and the cost of that archive does not appear on any invoice. It appears as a marketing department that spends its Thursdays on tickets, a search presence diluted by pages nobody defends, and a generative answer quoting a price you stopped charging in 2022.

The fix is not a better vendor. It is a rule set, an owner and a recurring calendar entry, all three agreed before the next model year lands.