There is a sentence on your website that has been wrong for a month. You know which one. Maybe it is a price. Maybe it is a product that launched in August and still says “coming soon.” Maybe it is the team page with someone who left in the spring.
You have not fixed it because fixing it is not a thirty-second job. It is a ticket. The ticket goes to a developer, or to the agency, and it waits behind other work, and at some point you stopped noticing the sentence because noticing it was annoying.
We have been building websites since 1999. For sixteen of those years we built them on HubSpot, and before that on everything else. That wrong sentence is the first of five signs that a company has outgrown its website, and the five show up on every platform in roughly the same order. Which is why “have we outgrown WordPress?” or “have we outgrown Webflow?” is the wrong question. You do not outgrow a platform. You outgrow a website.
How do you know you have outgrown your website?
Five signs. A routine change needs a developer and takes weeks. Only one person can safely edit the site. You pay for tools because the site cannot do something on its own. It was fast at launch and is slow now. And when your buyers ask ChatGPT about what you do, someone else gets quoted.
Each one gets taken apart below: what it looks like, what it is quietly costing, and what fixes it. At the end there is a five-question score that turns the signs into a decision, with what to do at each level.
Sign one: a routine change takes weeks
The wrong sentence is one instance. The pattern behind it is that on most websites the content is wired into the page. Changing the content means changing the page, changing the page means touching the template, and touching the template means finding the one person who understands the template. On every platform we have built on, that person is scarce and busy, and marketing is last in their queue.
So the request waits. A week for a price. Two for a new office location. Three for the landing page, which ships the week after the campaign it was for.
The cost is not developer hours. It is the requests that stop getting made. After the third three-week ticket, marketing quits asking. The campaign launches without its landing page. The claim that is no longer true stays up. The site drifts away from the business one unmade request at a time, and nobody puts that on a report because there is nothing to measure. It is the absence of things.
What fixes it. Separate the content from the page. When the pricing table is a file that says what the prices are, and the page is a template that reads that file, changing a price is editing a file. Nobody needs to understand the template. That is the whole idea behind content-first rendering, and it is the difference between a headless CMS and a page builder.
Here is what it looks like on the sites we run now. A home builder’s marketing coordinator told us on a call last week, “I update our website. I know what I want to do. I just want to do it.” So she types the change into the site’s agent in plain English. The agent makes it on a draft copy of the site, she looks at it, she says go live. Five to ten minutes, start to finish. Her colleague asked whether their requests would sit in a queue waiting for us to approve them. They do not. The only queue is if two of her own teammates ask for something at the same moment.
If a change to your site takes longer than it takes to describe the change, the site is the problem.
Sign two: only one person can safely edit the site
Every site has someone who knows where the bodies are. On Webflow it is the designer who built the class structure. On WordPress it is whoever set up the page builder and the plugins that make it behave. On HubSpot CMS it is the developer who wrote the modules. The site works because they are there.
Now ask what happens when they are not. On holiday. On another project. At another company.
At most mid-market firms the honest answer is that nobody else touches the site until they are back. That is key-person risk sitting inside the single most visible asset the company owns, and it does not show up on any org chart. We have watched the Webflow version of this often enough to predict it: by year three a mature site tends to run through one person who understands which parts are load-bearing. That person is usually excellent. They are also a single point of failure.
Often the one person is not a person. It is the agency that built the site, on a monthly retainer, and the retainer exists because nobody inside the company can make a change without them. We have been that agency. The retainer is real work and it is honestly billed, and it is also the clearest sign that the site belongs to the agency more than it belongs to you. The Five Locks piece calls this the dependency lock, and it is the one companies notice last because the invoice looks like a service, not a constraint.
The platform is the least useful part of the diagnosis. The five signs show up on every one of them, in the same order.
The quieter cost is that everything routes through that one person or that one agency, so sign one gets worse every year. A software company we audited last month had contractors signed up to the platform “for web stuff,” each one a paid seat, each one a person who knew a corner of the site and nobody else did.
What fixes it. Put the structure in the open. Content in files anyone on the team can read. Templates in a repository with a history of every change and a way to undo any of them. A draft site where changes are previewed before anything goes live.
When a site is built so that mistakes are visible and reversible, more people can be trusted to change it. On the home builder’s site, two people in marketing make changes, tag each other for a second look, and publish. On an entertainment client’s site, three people do. The expert stops being a bottleneck and goes back to being an expert. And the agency, if you keep one, is working on the next thing rather than being paid to be the only one who can touch the current thing. We wrote about how that works when the routine maintenance is an agent’s job in An AI Workforce for Your Website.
Sign three: you pay for a tool because the site cannot do the thing
Count the subscriptions that exist only to work around the website.
The software company we audited was running 90 live workflows, 65 forms and 196 reports behind one marketing site. Most of the forms were duplicates. A connector tool was copying blog posts from a shared drive into the CMS. A freelancer was rebuilding the reports in a separate dashboard product because the platform’s own reporting was, in the marketing lead’s words, “pretty bad for our tier.” Nobody built that on purpose. It accumulated, one reasonable decision at a time.
Here is the usual list. Each tool was a sensible choice on the day it was bought. Each one is also a monthly fee, a script on every page, a login somebody has to remember, and one more thing that can break on a Tuesday.
| What you pay for | Examples | Rough list price | On a site built properly |
|---|---|---|---|
| A form builder | Gravity Forms, Typeform, HubSpot forms embedded on a non-HubSpot site | $59 to $259 a year for a plugin; $25 to $50 a month and up for a hosted builder | Forms are part of the site. They post wherever you say, save a copy you own, and notify the right person. No embed, no seat |
| A survey tool wired into the CRM | Paperform, SurveyMonkey, Typeform again | $20 to $100 a month | The survey is a page. Answers land in your CRM and your own database directly |
| A scheduling embed | Calendly, HubSpot Meetings | $10 to $16 per seat per month | The site books the meeting, creates the calendar event with the video link, and runs the follow-up that meeting type needs |
| A connector or automation tool | Zapier, Make | $20 to $30 a month to start, $100 and up for a team | The site owns the integration. When a form is submitted, the code that runs next lives in the website |
| A site search add-on | Algolia, SearchWP | Usage-based, or $99 and up a year for a plugin | Search runs over your own content, all of it, as part of the site |
| A pop-up or lead-capture tool | OptinMonster, Hotjar’s survey widgets | $7 to $147 a month | Personalisation is logic in the site, reading your own visitor data |
| A second analytics or heat-map product | Hotjar (now Contentsquare), Looker Studio dashboards built by a freelancer | $49 a month and up, plus the freelancer | First-party analytics on your own site, with the data in your hands and per-page speed in the same dashboard |
| An image optimisation plugin | Smush, ShortPixel | $50 to $100 a year | The build optimises every image at publish. There is nothing to install |
| A security or backup plugin | Wordfence, UpdraftPlus | $70 to $149 a year, each | The site is a repository. Every version is a backup. There is no plugin layer to patch |
Add it up for a typical mid-market site and the workaround layer alone runs from a few hundred to a few thousand dollars a year on list prices. That is before the hours. Mid-market companies run 12 to 20 marketing tools on average, 2 Source 2 Digital Bloom, cited in GTM8020, 38 Marketing Technology Stack Statistics Every CMO Needs in 2026. The same article notes 92% of organisations keep their stacks under 20 tools. and one 2026 survey of 2,400 marketers found that only about a third of the tools in a marketing stack are actively used in any given month. 3 Source 3 Visionary Marketing, MarTech Stack Study 2026: 34.1% of marketing stack tools actively used in a given month. Single-source survey; treat the exact figure with care, the direction is consistent across studies. The rest are paid for, loaded on the page, and forgotten.
The money is rarely in one budget line, which is why nobody adds it up. If you want the full arithmetic for one platform, How Much Does HubSpot Actually Cost does it line by line for HubSpot, and the Full-Stack Cost Index is the framework for doing it yourself on any platform. The less obvious cost is that every one of those tools is a script, and scripts are what make a site slow. That is sign four.
What fixes it. A site that can do the thing. Read the right-hand column again. None of it is exotic. It is what a website is when it is an application the company owns rather than a page builder someone rents. In the audit, the recommendation for most of the 90 workflows was the same: they go away, because the thing they were working around does not exist on the new site. The argument that your CMS is the first subscription worth replacing is in Your CMS Is a SaaS Product Too.
Sign four: it was fast at launch and slow now
Nobody launches a slow website. Sites get slow.
The marketing team adds the chat widget. Then the heat-map tool. Then the video embed on the home page, fonts from a second provider, a tag manager with forty tags inside it, and the hero image someone uploaded at full resolution because the platform promised to optimise it. Eighteen months later the page that scored well at launch takes four seconds on a phone, and the person from sign two is asked to find out why.
This is the sign companies measure most and understand least. A speed score is a symptom. The disease is a platform that lets anyone add anything to every page, and never removes anything.
The cost is conversion, because most of your buyers meet you on a phone and a measurable share of them leave before a slow page finishes loading. There is a second cost arriving now. The LLMs your buyers ask are reading your pages too, and a page that renders its content late, behind scripts, is a page they often do not read at all. That argument is in Why Content-First Rendering Matters, and the wider case for the frameworks that render that way is in Web Frameworks for Marketing Websites.
What fixes it. A build that renders the content first and ships only the scripts a page needs. On the sites we run, every page has a speed score in the dashboard the client sees, so an entertainment company can look at its gallery page, see that it is heavier than the rest because there are two hundred images on it, and decide whether that is fine. Nobody has to guess. And the chat widget can still exist. It just cannot slow down the pricing page, because it was never loaded on the pricing page.
Sign five: buyers ask ChatGPT about you, and someone else answers
This is the newest sign and the one most teams have not checked.
Open ChatGPT, Claude, Perplexity or Gemini and ask the question your buyers ask before they call you. How long does a migration take. What does a custom home cost per square foot in your county. Which firm handles the thing you handle. Then look at who gets quoted.
If it is a competitor, or a directory, or a forum thread from 2019, your site is not being read. Usually that is because of signs one through four. Content buried in a page builder, rendered late, behind scripts, with no structure that tells a machine what is a product and what is a case study, is content an LLM cannot use. Your Website Is Invisible to AI covers why that happens and what it is costing. The short version is that LLMs quote what they can read, and most sites are not readable.
The cost is a buyer who never arrives. They asked, they got an answer, the answer named somebody else, and they called somebody else.
What fixes it. Two things, and they are the same two things as the other four signs. Content as structured files, so a machine knows what each page is. And a site that renders that content first, so a machine can read it. AEO is an architecture problem before it is a marketing one, which is the good news, because architecture is fixable once.
On a site built that way, this sign turns into a job. The discoverability agent we are building for our clients’ sites watches the questions buyers ask, who gets quoted, and what a competitor answers that you do not, and hands the gap to the content agent as a brief. A client with a book launch gave us a small version of what that site awareness looks like last month. He asked the site agent to update the release date on the one page he knew about. The agent came back and said the date was on six other pages, and asked whether he wanted those changed too. That is what it looks like when a site knows what it contains. A page builder cannot do that. It does not know what a release date is.
Score your website
Five questions, one per sign, each scored from 1 to 5. The total is out of 25, and the result tells you what to do next, not just where you stand.
Then ask marketing one more question, which is not on the card: which request did you stop making because the answer was always a ticket?
The answer is usually specific. A landing page per campaign. A pricing page that updates when the product does. A calculator on the pricing page. A resource library someone could actually filter. Whatever it is, it has been costing you longer than any of the five signs, because somewhere along the way it stopped being a request and became an assumption about what the website cannot do.
What you get when it is fixed
A website that keeps up with the business instead of trailing it. Changes ship the week they are decided, from a preview, by the people who asked for them. Anyone on the team can be trusted to make them, because every change is visible and reversible. The workaround subscriptions quietly stop renewing. Pages stay fast because nothing can slow them down by accident. And the content is files the company owns, which means an AI agent can maintain the site and ChatGPT can quote it.
You will also never do this again. Once the content is yours and the site is built to be changed, the next redesign is a design project, not a migration. The site goes through as many iterations as the business needs on the same foundation. The Sovereign Stack Blueprint describes that foundation; What Is an AI-Native Website makes the case for owning it. And if you are on HubSpot, What Happens to Your Content When You Leave HubSpot is the migration guide we wrote after moving our own site.
Do the count before the renewal, not after. Twenty or more means move. It is a date your vendor already has in their calendar.
The research is free. So is the diagnosis. 60 seconds →