Every comparison of these three lines up templates, apps and SEO panels — things you evaluate on the afternoon you buy, none of which is what you will care about in two years.
Two things decide this properly. Whether a machine can read you, and what it costs to change your mind. Both have moved sharply in the last eighteen months, and most comparison articles have not caught up.
The gap nobody put on the chart
Testing through 2026 established something that changes how this decision should be made:
GPTBot, ClaudeBot and PerplexityBot do not execute JavaScript. They do not wait for rendering, and they do not make a second attempt. GPTBot was observed fetching JavaScript files in about 11.5% of requests without ever executing them; ClaudeBot downloaded them in about 23.8% and executed none.
Googlebot is the main exception — it renders fully, and Gemini borrows that rendering.
Read those two paragraphs together and the consequence is uncomfortable: your site can rank perfectly well on Google and be effectively invisible to ChatGPT, Claude and Perplexity. Not ranked low. Absent. The content those assistants would have quoted was drawn by JavaScript a fraction of a second after they had already left.
There is a ten-second test, and it is worth doing before reading further. Open your site, view source — the raw HTML, not the element inspector — and search for the sentence describing your main service. If it is there, AI crawlers can read you. If it only shows up in the inspector, it was rendered client-side, and a large share of AI retrieval never saw it.
And on a builder, you do not control the layer that fixes it
This is where the platform choice stops being about design.
Squarespace does not let you edit robots.txt at all. Not restricted — absent. Which matters more than it sounds, because of how the file works: a crawler that finds a group matching its own user-agent ignores the general * group entirely. Naming OAI-SearchBot, ClaudeBot and PerplexityBot explicitly is how you make access a decision rather than an accident, and on Squarespace you cannot make it.
Wix does let you edit robots.txt, through its SEO dashboard — genuinely better. But Wix does not allow arbitrary files at your domain root. That is a platform restriction rather than a setting anyone missed: files uploaded through its media manager land behind a CDN path, not at yourdomain.com/.
Squarespace's structured data is rigid and cannot be edited. This is the one that quietly costs the most. Structured data is where you state, in the form a machine reads, exactly what you do, where you do it, and what proves it — and whether an assistant can name you depends on those facts being stated precisely. A platform that will not let you correct how it describes your business has made that decision permanently, on your behalf.
None of this is an argument that builders are bad. It is an argument about who holds the controls, and on a builder the answer is not you.
Speed stopped being a nicety and became an entry condition
The old framing was that faster sites rank slightly better. That framing is now too weak.
Those AI crawlers do not wait for rendering and do not make second attempts. A slow response, a timeout, a server having a bad minute — there is no retry queue and no second chance. Speed and reliability are not ranking factors here. They are inclusion conditions.
Which makes architecture, rather than tuning, the thing that matters:
| Typical page load | Notes | |
|---|---|---|
| Prerendered (static) | 0.5–1.2s | No per-request work; the page was finished before the visit |
| Default WordPress | 2.5–4.0s | PHP and a database assemble the page while the visitor waits |
Among the builders, Wix passes Core Web Vitals on about 74% of mobile pages against Squarespace's 70%, and the median mobile Lighthouse score is 62 against 30. Squarespace can be pushed over the line with disciplined image compression and restraint about embeds, but that is ongoing work rather than a property of the platform. Neither leads the category — Duda passes at about 85%.
Read those as medians across millions of sites: they describe what each platform produces when nobody is paying attention, not a ceiling. A maintained site on a slower platform beats a neglected one on a faster platform.
Custom sits at both ends of that range, which is why it is missing from most charts. Built carelessly it is the slowest option available, because nobody is stopping you. Built as prerendered HTML it is the fastest, for a structural reason rather than an optimisation one: there is no per-request work to be slow at.
Not all custom is the same, and this gap is wider than the builder gap
This is the part almost nobody explains before you commit, and it matters more than the Wix-versus-Squarespace question you arrived with.
"Custom" covers two genuinely different things.
A CMS build, usually WordPress. Every visit runs PHP, queries a database and assembles the page while the visitor waits. Plugins add features and add work.
A prerendered build. Pages are generated once, at build time, and served as finished HTML. No database call during a visit, because the page already exists.
Three consequences follow, and they show up in numbers rather than opinions.
The content is in the HTML by construction. Not as a fix, not as a plugin, not as a rendering workaround — it is simply how the page is made. Given what the crawler research says, that is no longer a technical preference. It is the difference between being quotable and being skipped.
Speed is a default rather than a project. Around 44% of WordPress sites pass Core Web Vitals on mobile, and only 32% have good server response times. The usual cause is hosting: WordPress on shared hosting averages a p75 TTFB of 900–1,400ms, and the same site on managed hosting with server-level caching drops to 120–250ms. That fix is real and available — it is also a recurring bill and a standing maintenance commitment. A prerendered site starts where the optimised WordPress site finishes and stays there untended.
The security surface is largely absent. WordPress core is reasonably hardened; the plugins are the problem. Trackers logged 253 new plugin vulnerabilities in a single week of January 2026, roughly 36 plugin disclosures a day across the year, and 95% of WordPress compromises begin with a vulnerable or compromised third-party plugin. Two details make that worse than the headline: 43% require no login to exploit, and 23% had no patch within 30 days of disclosure — so an owner who updates diligently every week is still exposed to about a quarter of them. A site with no plugin layer and no database does not have this category of problem at all.
Where WordPress genuinely wins: a large editorial team publishing daily, a real e-commerce catalogue, or a business that depends on a specific mature plugin. Anyone who tells you WordPress is simply the wrong answer is selling something. The honest position is that its strengths are ones most small service businesses never use, while its costs — hosting quality, update discipline, plugin risk — they pay in full.
What it costs to leave
The other decision that outlives the build.
| Getting your content out | Getting the site out | |
|---|---|---|
| Wix | Text, images and store data can be retrieved | Not possible. Wix's own documentation states a site cannot be exported or hosted elsewhere |
| Squarespace | Older versions export XML; 7.1 has no XML export | Design and layout do not transfer; the front end depends on Squarespace's own CSS and JavaScript |
| Custom build | Everything, always | Already yours — files, database, hosting account |
Wix's restriction is architectural rather than commercial: the site runs on proprietary technology, so there is nothing standard to hand over. Leaving Wix is never a migration — it is a rebuild, at rebuild prices, in a year you had not planned for one. Squarespace has moved quietly in the same direction; a site built on 7.1 today is far closer to Wix's position than its reputation suggests.
One thing worth doing regardless of what you choose: keep your own copy of your content outside the platform — every page's text, every photograph at full resolution, every testimonial with its attribution. It costs an afternoon, and it is the difference between leaving for the price of a rebuild and also paying to reconstruct four years of writing from screenshots.
What to ask whoever builds it
Custom only delivers the advantages above if it is built that way. These five questions separate a modern build from an old one wearing new words, and they work on us as well as on anyone else:
- Will my main service copy be in the HTML source, or drawn by JavaScript? There is only one right answer now, and you can verify it yourself on the day it launches.
- Is the structured data written for my actual business, or generated from a template? Generic markup describes a generic business, which is what an assistant will then repeat.
- Is search and AI readability part of the build, or a separate package afterwards? This is where quotes diverge most. Structure decided at build time costs nothing extra; the same structure retrofitted later is a second project, priced as one, working against decisions already made.
- What is the ongoing security and update commitment? A plugin-based build carries a permanent maintenance obligation. A prerendered one largely does not. That difference compounds over years.
- On the day I leave, what do I walk away with? Domain, code, content, hosting and analytics accounts should be in your business's name from the first day, not transferred as a favour at the end.
Point three is the one most worth pressing. The common pattern is a build quoted at one price and "SEO" sold afterwards as a retainer — which is more expensive in total and less effective, because the structural decisions that determine machine readability were already made and are now being worked around. Our own builds include the crawlable structure and the business facts in machine-readable form as part of the build, because retrofitting them is the more expensive way to arrive at a worse result.
When a builder is genuinely the right answer
Worth stating plainly, because it is often true.
If the site's job is to look real when somebody checks you out before calling, if your work arrives entirely through referral, and if nobody is going to be searching your category and choosing between you and a competitor — a builder does that job completely, for $17–$99 a month, starting this afternoon. Prefer Wix if control and performance defaults matter more; prefer Squarespace if design consistency matters more and nobody on your side has an eye for layout.
The decision flips when you need to be found and chosen rather than merely confirmed. That is the point where not controlling robots.txt, not controlling your structured data, not being able to guarantee your content is in the HTML, and not being able to leave stop being abstractions and start being the reason a competitor is named in an answer and you are not.
Our Starter Website is $695, one-time: one designed page, copy written from facts you confirm, your services and service area stated properly, a working inquiry path, and the crawlable structure and machine-readable business information done at build time rather than sold back to you later. Launched on a domain registered in your name, with everything yours from the first day.
If you already have a site and want to know whether its problem is the platform or the content on it, the free AI Search Snapshot answers that specific question before you spend anything. Deeper work starts with the Upgrade Blueprint at $299, fully credited.
On price, the honest bands for each route are in what a small business website actually costs.
Sources
- Testing on AI crawler JavaScript execution — GPTBot, ClaudeBot and PerplexityBot not executing JavaScript, the ~11.5% and ~23.8% fetch-without-execute figures, and Googlebot as the rendering exception — is drawn from 2026 server-log and crawler-behaviour analyses published across the search-industry press, including SearchOptimo's crawler rendering tests. These are practitioner measurements rather than vendor documentation; the underlying behaviour is straightforward to verify yourself with the view-source test described above.
- Wix — Editing your site's robots.txt file and Exporting or embedding your Wix site elsewhere. Wix's own documentation on robots.txt editing and on why a site cannot be exported.
- Squarespace — Exporting your site. The robots.txt limitation is documented in Squarespace's own support forums, where the absence of an editing method is confirmed repeatedly.
- Core Web Vitals pass rates and median Lighthouse scores for Wix, Squarespace and Duda, and the WordPress pass rate, TTFB bands and static-versus-CMS load times, come from 2026 platform performance comparisons compiled from public field data. They are medians across large populations, not ceilings for individual sites.
- WordPress plugin vulnerability figures — weekly disclosure counts, the 95% breach-entry share, and the 43% unauthenticated and 23% unpatched-within-30-days proportions — are from 2026 WordPress ecosystem security reporting.
- Builder subscription pricing ($17–$99 a month depending on plan) is from Wix's and Squarespace's own published pricing pages.
- LifeVen's own price is published on the Starter Website page.
Checked 2026-07-31. Platform capabilities and crawler behaviour in this area change quickly; the view-source test is the part that will still be true whenever you read this.