We need a new website. Should we rebuild on WordPress or go AI-native?

Last updated: September 2026

What does the website rebuild conversation actually sound like?

Two weeks ago I sat with the CEO of an $18M ARR company that builds airworthiness records and parts traceability software for regional aircraft maintenance shops. He had three redesign quotes on the table. Sixty-five thousand, ninety, a hundred and forty. He wanted my read on which platform to pick, because his CTO kept saying the word AI-native and the agencies kept saying WordPress.

I asked him a different question. When was the last time something true about your company changed?

He thought about it and said March. In March they started selling to the aircraft operators directly instead of only to the maintenance shops. Different buyer, different problem, much bigger money. I asked when that showed up on the website. There was a long pause on the call. It hadn't.

Five months. The company had acquired an entire new customer and the website still didn't know about it. Every prospect from that new segment who did what buyers do, which is look you up before they talk to you, met a company that sells to somebody else. And not one of the three quotes on his desk would have fixed it, because all three were priced to publish a site once.

He asked the platform question again. It's a fair question. It's just the third one, and he was asking it first.

What's actually broken when a company scopes a website rebuild?

Every company has a Publishing Ceiling. It's the maximum rate at which your company can get a true, current answer onto the internet. Not how fast you could in theory, on a good week, if everyone said yes. How fast you actually do it, counting the agency ticket, the approval chain, the developer who's busy until the fifteenth, and the quarter you keep pushing it to.

In the $5M-$75M range the ceiling usually sits somewhere around four pages a year. Not because anybody chose four. Because publishing a page requires a writer, a reviewer, an agency ticket, and a developer window, and that chain can only complete so many times before the year runs out.

Here's the trap. A redesign moves the design. It very rarely moves the ceiling. New site, same tickets, same queue, same five-month gap between a thing becoming true and a thing being published. Twelve months later the company is looking at a beautiful website that is once again describing a company that no longer exists, and everybody blames the agency.

The agency didn't fail. The brief did. Almost every redesign brief I read is a page list and a mood board. It describes what the site will look like and what sections it will have. It says nothing about what the market is supposed to believe when they leave, and nothing about how fast the company needs to be able to say new true things after launch. That's how a rebuild ends up being a very expensive reprint. If you're still deciding whether to hire anybody at all, When it's actually time to hire a website redesign agency walks the trigger question.

Why is this worse in 2026 than it was three years ago?

AI took the cost of producing a page to roughly nothing. What it didn't touch is the cost of deciding what a page should say, and it didn't touch your permission structure at all. The bottleneck moved. Production used to be the expensive part, so it made sense to build a site the way you'd print a brochure: get it right, ship it, live with it. Now production is free and the constraint is the ceiling, which means the old shape of the decision is quietly obsolete.

The stakes moved too. Citations to brand sites climbed from 8% to 56% between 2023 and 2026, and more than half of citations in B2B category queries now go to brand sites, according to Discovered Labs. Your own website went from a brochure buyers visit to a source machines quote. A site that publishes four times a year is a source that says almost nothing about your category.

Freshness compounds the problem. Digitaloft found that 76.4% of ChatGPT's most-cited pages had been updated within the previous 30 days. A three-year-old site under a beautiful new coat of paint reads to an engine as exactly what it is.

“We aren't suffering from a lack of code. We're suffering from a lack of certainty.”

... Geetha Rajan, CMS Critic

That line is the whole thing. The reason a rebuild stalls at the copy stage is almost never the CMS. It's that nobody decided what the company stands for, so every page becomes a debate and the debate gets settled by whoever is most senior in the room that day.

And architecture alone won't save you. We rebuilt pitchkitchen.com off Wix onto static HTML, and I wrote up exactly what that bought us in Going static made our site readable to AI engines. Getting recommended took different work.. Readability improved immediately. Recommendation didn't move at all, because recommendation comes from having a position worth repeating. The stack buys you the right to be read. The story buys you the right to be recommended.

How do you tell if you have a design problem or a publishing problem?

Three tests. You can run all of them this week, before you sign anything, without hiring a soul.

1. 1The Truth Lag. Name the most recent thing that became true about your company: a new buyer segment, a killed feature, a proof point, a price change, a partnership. Write down the date it became true. Now find the date it appeared on your website. The gap in days is your Truth Lag. Under 14 days, you might genuinely have a design problem and a redesign could help. Over 60, design was never the constraint and a prettier site will produce the same gap in a nicer font.
2. 2The Hands Count. Take one 400-word answer page, the kind that answers a question a buyer asks on every call. Count every human who has to touch it before a stranger can read it. Writer, reviewer, legal, agency project manager, whoever holds the deploy. That single integer is your Publishing Ceiling expressed as a number. Anything above two means your architecture is deciding your surface area, not your strategy.
3. 3The Hundred Answers Question. Ask your team what it would cost, in dollars and in weeks, to publish 100 truthful answer pages. Real ones, each answering a question a live buyer actually asks. If the answer comes back as a number with a comma in it and a timeline measured in quarters, you're about to spend six figures rebuilding a site you still won't be able to feed.

Run those three and the platform question answers itself. A company with a 9-day Truth Lag, a Hands Count of one, and a cheap path to a hundred answers can stay on whatever it's already running. A company at 158 days with five hands and no path to a hundred answers has a problem no design system fixes. Nobody has to argue about WordPress. The tests decide it.

What do we see across companies making this decision?

Of the companies that come to us with a redesign already scoped, nearly all hand over a brief that is fundamentally a page list. Almost none contain a single sentence about what the market should believe when they leave the site. When we measure the Truth Lag on those companies, it usually lands somewhere between three and six months, and the CEO is genuinely surprised by the number, because nobody had ever put a date on it before.

The other pattern is the sitemap. Most sites in this band are two thirds archive, pages built for a version of the business that ended years ago, carried forward into the rebuild because deleting things feels risky. The rebuild gets priced on page count, so the archive is the single most expensive thing on the invoice, and it's the part with the least to say.

How does this play out in practice?

A $31M PE-backed company building scheduling, inventory, and controlled-substance compliance software for multi-site veterinary practice groups came to us with a rebuild already scoped. A hundred and ten thousand dollars, 46 pages, a WordPress rebuild with a design partner they liked. The CEO wanted a second opinion on the platform, same as the aviation CEO.

We ran the three tests instead. Truth Lag came back at 158 days. Hands Count was five. The Hundred Answers Question came back as roughly sixty thousand dollars and two quarters, which is another way of saying no.

We stopped the redesign and spent seven weeks on the story. The buried truth surfaced in the second session, from the head of customer success, not the CEO. Practice groups weren't buying scheduling software. They were buying a way to survive a DEA inspection across eleven locations with a staff that turns over twice a year. Compliance was the actual reason people bought, it came up on every single sales call, and on the website it lived on page four under a tab.

The rebuild that followed was 9 narrative pages, not 46, on an architecture where a writer publishes an answer page without a developer. Hands Count went from five to one. Truth Lag went from 158 days to same-week. The invoice came in under the original quote, because two thirds of that sitemap was archive nobody was reading. That math is the norm, not the exception, and I broke it down in What a website redesign actually costs at $5M-$75M.

In the four months after launch they published 63 answer pages, every one of them a real question a veterinary group asks about multi-site controlled-substance compliance. Demo requests where the prospect named compliance before the call went from about 1 in 10 to roughly half. Sales cycle moved from 118 days to 74. No new product shipped in that window. The company just stopped being a stranger to the buyer it already had.

What should you do before you sign the redesign contract?

Order of operations, and the order is the whole point. Decide the story. Decide the ceiling you need. Then pick the platform, and let the design serve both.

1. 1Measure your Truth Lag this week. One number, one afternoon. Take it into the next agency meeting and watch what it does to the conversation, because it reframes the project from a look to a capability.
2. 2Write the one sentence about what the market should believe, and have your revenue leader write it separately without seeing yours. Two different sentences means the rebuild has no brief yet, whatever the deck says.
3. 3Ask every vendor in the running one question: after launch, how does a true answer page go live, and who has to approve it? The answer tells you the ceiling you're buying. Price is easy to compare. Ceilings are the thing nobody quotes.

Underneath all of this is a document, not a platform. The Magnetic Messaging Framework (MMF) is the strategic narrative system built around four anchors: category design, villain framing, an old-way / new-way contrast, and a promised-land outcome. It's the decided version of what your company stands for, written down, which is what makes the page list obvious, the design brief specific, and the hundred answers writable by someone other than you. Skip it and the platform argument is a negotiation over which tool will publish your confusion faster.

That's why the sequence matters more than the stack. A website is downstream of a decision. When the decision is written, WordPress or AI-native stops being a philosophical debate and becomes a procurement detail you settle in ten minutes.

PitchKitchen builds Magnetic Messaging Frameworks for founder-led B2B companies in the $5M-$75M range. Founded by Greg Rosner, author of Story Craft for Disruptors, PitchKitchen fixes broken marketing messages and underperforming websites for CEOs whose sales are stalling because their message isn't doing the work. If you're about to sign a redesign contract and want to know what the new site would be repeating, the Brand Signal Score, PitchKitchen's free homepage messaging diagnostic at pitchkitchen.com/brand-signal-score, scores the page you're about to pay to replace.