Landingsite.ai test data

Each of the three fixed briefs was typed word for word into Landingsite's two boxes: a business name and a business description. The clock ran from pressing Create Your Website to the moment every page in the brief opened at the site's preview address. The free plan cannot publish, so a live address was never reached.

The first look at the cafe site came about 20 seconds after I pressed the button. The AI then kept working. It stopped about 15 minutes 27 seconds after the brief, because it spent about 10 minutes chasing photo boxes it thought were blank. They were not: when I scrolled the page in a real browser the photos loaded. The photographer's home page was done in 1 minute 40 seconds and the shop's in about 4 minutes.

The AI builds the home page first and links to the other pages. Each site needed one more message, 'Build the Menu, About, Catering and Contact pages now', which cost 1 free chat on the cafe and 2 on each of the others. With that message the five pages of the photographer site were ready after about 4 minutes, the six pages of the shop after about 9 minutes 40 seconds, and the cafe after about 27 minutes (22 minutes of AI work and a 5-minute gap before I sent the pages message).

The form asks only two questions. Both are required. A third, optional question, 'Where did you hear about us?', appears while the site builds and has a Skip link. The name and the type of business were on the draft every time (3 of 3). Landingsite also asked for no logo or photos: the AI makes its own logo and picks its own stock photos.

• Times were read from my own session log. The 5-minute gap on the cafe was mine.

The twelve points were read in the HTML of the first drafts' preview address. The sitemap, robots.txt and a per-page noindex cannot be read before a site is published, so the test is not graded.

6 of 12 points were there: an editable title and description per page (the AI wrote them and rewrote three on request), one H1 on every page, alt text on every image, clean page addresses, and a Redirects screen with 301 and 302 redirects and wildcards.

3 were missing on all 16 pages: a canonical tag, structured data and social-card (Open Graph) tags. The sitemap and robots.txt gave a 'Preview not found' page, so I could not read them; a per-page noindex came only from asking the chat (it set one on the password page).

The Spanish version declares both languages with hreflang tags. The cafe's descriptions ran 152 to 165 characters, over the 155 limit on four of its five pages. The builder has no SEO checklist; asking the AI for titles and descriptions got 2 of 3 inside the length limits (WB-S3).

Lighthouse scored SEO 63 on every page because the preview is marked noindex; nothing else failed. Google Search Console and Bing (WB-S1) and the AI-crawler rules (WB-S4) need a published site, so they were not run.

• Everything here is the preview address; a published site may add a sitemap, robots.txt or tags.

Each of the twelve points and the added business checks was tried as far as the free plan allowed. At each paywall I recorded its words and the price to pass it.

Blocked by a paid plan: a custom domain (the Domain screen says 'Domains are available on paid plans'), a test order and taxes (the store says 'Your store can't take payments yet. You've added something to sell, but taking payments (and publishing your site) needs a paid plan'), a mailbox on a domain, and the monthly report ('Monthly reports go out for published sites on a paid plan'). Selling needs Pro at $20 a month billed yearly, with a 1% fee on top of Stripe's.

A discount code, stock count and abandoned-cart email (WB-F3): 0 of 5 worked. The AI said discount codes need a connected Stripe account, which I did not create. It set stock to 5 and added a 'Sold out' state but I could not order to test it. It said it cannot send abandoned-cart emails at all.

Worked on the free plan: a contact form with Turnstile spam protection, a submissions list and webhooks (2 of 4: I found no CSV export, auto-reply or file-upload field; WB-F2); a blog with one post, a category and two tags (1 of 4 for WB-F1: the RSS address returns a web page and not a feed, and there is no scheduling or comment setting); a Spanish version of the whole site with a switcher and hreflang (3 of 4; the cost is not shown; WB-F4); a logo tool (WB-A4: the logo has the right name, the download needs a paid site or a $5 purchase).

A members-only page (WB-F5) works as a soft gate: a password prompt is built into the page, and the password can be read in the page source, because the site is static files. Roles for a collaborator (WB-C1): Admin, Editor and Read-only; I did not invite anyone. A headless CMS with collections (WB-F8): not offered in settings. Integrations (WB-C3): not run; the form settings have a webhook for Zapier or Make.

• No domain was connected, no order was placed and no mailbox was made.

I looked for an export in the builder, in the site settings and on the help page, read the terms of service and the pricing page's refund and cancel answers, and opened the billing page.

I found no way to export the site's code, pages or images. The Ownership screen in each site's settings is blank on the free plan. The help page's only answer near the subject is about moving a domain.

The terms (last revised 11 July 2023) say fees are non-refundable except as the agreement allows, that a subscription renews unless cancelled at least one day before, that Landingsite is not obliged to back up your content, and that your content may be deleted at any time without notice. The pricing page says there is a 30-day money-back guarantee and that you cancel in Settings and billing, after which the site stays active to the end of the period.

The cancel path cannot be tested on a free plan: Manage Billing opens a Stripe page that says 'No payment method' and 'No invoice history', with no cancel button (WB-O1).

Restore an earlier version (WB-R1): every AI message and every text edit is saved as a version, 14 on the cafe. I rolled back to an earlier one, confirmed a warning ('Any uncommitted changes in the editor will be lost') and the new version appeared about 10 seconds later. The version's own preview address showed the old page when I checked a few minutes later, but the main preview address still showed the newer page 50 minutes after the rollback.

Security (WB-R3): the preview answers with 0 of the 6 headers I check (Strict-Transport-Security, Content-Security-Policy, X-Frame-Options, X-Content-Type-Options, Referrer-Policy, Permissions-Policy), http is not redirected to https, and the certificate is a valid Google Trust Services wildcard for the preview domain. The account page offers a password and a list of active devices, and no two-factor sign-in.

Cookies and privacy (WB-R4): the AI wrote a cookie banner that only shows where the law needs one, with a cookie settings link, and a privacy page; a clean visit to the shop set 0 cookies before consent.

• Terms and pricing text read on 28 October 2026.

I emailed Landingsite support seven questions on 28 October 2026 (the thread is logged in the evidence folder) and read the help page. I did not search my inbox for replies.

No reply had arrived when I wrote this. Support is not graded. The help page says the team 'typically respond within 24 hours'.

The help centre is one FAQ page with no search box. Five searches (WB-H1): 4 of 5 questions were answered correctly (connect a domain, form messages not arriving, change a card, add custom code through the AI); export my site was not answered.

The AI chat is the first line of support: the help page says to ask the AI first, and the questions I put to it about discount codes, abandoned carts and a password page got honest answers about what the site cannot do.

• Seven questions, sent once.

Every page of the three first drafts was loaded at 360, 768 and 1440 pixels wide on the preview address of the version that held all the pages (version previews of each site's first complete draft). The checker counted sideways scrolling, tap targets under 24 pixels, text under WCAG AA contrast (axe-core), placeholders, broken images and overlapping text, once per type per page. I then ran axe-core on every page at 1440 and 390 pixels, walked each home page with the Tab key, and compared the three drafts.

31 faults on 16 pages: 10 on the cafe (tap targets and pale text on all five pages), 12 on the photographer (tap targets on all five pages, pale text on two, and sideways scrolling on all five at tablet width) and 9 on the shop (tap targets on all six pages, pale text on two, and a 'Studio photo coming soon' placeholder on the home page).

The sideways scrolling is real: at 768 pixels the photographer's menu runs 13 pixels past the screen edge on every page. No broken image or overlapping text was found on the first drafts. The checker also flagged the 'Most booked' badge on the photographer's pricing page and a block of contact links on the shop, and I did not count either, because the badge is meant to sit on the card and the links stack correctly.

The AI kept reporting 'blank photo boxes' on the cafe. They were not blank: the photos load as you scroll, and its own screenshots had been taken before they loaded.

Look-alike check (WB-A2): 1 of 5 things were the same on all three sites. The hero has the same layout everywhere (text on the left, a tall photo with a floating card on the right). The cafe and the shop share a font pair (Fraunces and Karla), the photographer uses Cormorant Garamond and Jost. Palettes, section order and section styles differ, and I found no template name, so the three sites do not look like one template.

Accessibility (WB-D3): no critical problems on any page. Serious colour-contrast problems on 9 of the 16 pages (41 elements: 30 on the cafe, 7 on the photographer, 4 on the shop). Every image has alt text (0 missing). Walking the pages with the Tab key, I reached most links and buttons in order (30 of 39 on the cafe home page) and did not find why the rest were skipped; a focus ring was missing on 1 to 3 elements of some pages. The Spanish pages keep the page language set to English.

Switching template after the content is in (WB-D1): not possible. I found no template or theme picker in the builder or in settings; each site is generated fresh.

• The pages were checked on the preview address, not a published site. The tap-target and contrast counts follow the same rules as for Wix, Lovable and Hostinger. The cafe's hero sentence was my own change in the direct text editor, made before its other pages were built.

Each brief has 20 facts. Every page of each first draft was read and each fact marked right, missing or wrong. Anything the AI added that the brief did not give was listed as invented: policies, promises, figures and descriptions. Generic marketing wording was not counted.

All 60 of 60 facts were on the pages: 20 on the cafe, 20 on the photographer and 20 on the shop, with the right prices, hours, addresses, phone numbers and emails.

15 invented claims: 3 on the cafe (a 'call ahead and we'll have your order waiting' service, 'We reply during opening hours' and bread 'baked in-house daily'), 7 on the photographer ('Most booked' on the wedding package, 'Dates book out months ahead', 'The balance is due before the event', a studio 'in south Austin', a meeting-place process, 'Every session is led by me' next to a two-photographer package, and 'Every frame here was taken by Ines' beside stock photos) and 5 on the shop ('Each piece is wrapped by hand in the studio', product details such as a matte glaze and a ridged rim, the potter's process of throwing, trimming, glazing and firing, 'one kiln', and what the newsletter sends).

There were no invented reviews, awards, staff, prices or contact details. Every photo is a stock photo. The shop's three product photos are the text 'Studio photo coming soon'.

Logo and photos (WB-A3): the form has no place for a logo or photos, so I could not give my own. The AI made a logo with the right name on all three sites (3 of 3), and the free-plan chat has upload buttons for photos that I used once, on the shop.

• Claims count when they state something about the business the brief did not give. Words such as 'unhurried' and 'worth the walk' were treated as style.

The twelve fixed edits were done on the cafe site, one message each in Landingsite's chat, and checked on the preview address. The wording change and the footer line used the direct text editor in the preview, which costs no chat. When the cafe ran out of free chats, the two edits not yet done were tried on my other two sites. Five written instructions (WB-A5), a gallery, a video embed and a slider (WB-D5), free placement (WB-E2) and a site-wide snippet (WB-C2) were run as added checks.

10 of 12 edits worked on the cafe without a workaround: move a section (about 3 minutes), site font (1.5), site colours (3.5), a Gift Cards page with a menu link (1), a footer line (direct editor, free), a form field (1), a button link (1.5), a duplicated section (6), embedded HTML (1.5) and a rewritten section (1). The colour change replaced the green everywhere and kept the cream background.

The other two worked on my other sites, after the cafe's chats ran out: a phone-only change (the three numbers under the hero buttons are hidden on phones and shown on desktop) on the photographer, and a replaced hero image with its focal point set to the top left on the shop.

Each message cost one or two of the ten free chats, and the cafe hit the 'Chat Limit Reached' wall after 11 messages (one was lost when I reloaded the page mid-task).

Added checks: 4 of 5 written instructions did the job without a second try (a Gift Cards page, a Spanish version, a price-ordered list and descriptive alt text on every image; the opening-hours change I did not run). Gallery, video and slider (WB-D5): 2 of 3. The gallery opens a full-screen viewer and the slider works; no YouTube video appears on the page. Site-wide snippet (WB-C2): 0 of 2. The HTML embed showed up, but the tracking snippet is not in the page source of the three pages I read. Free placement and layering (WB-E2): no, there is no drag-and-drop editor, only chat and direct text editing. Median time per edit about 1.5 minutes (WB-E1); a blank page cannot be started, because the AI always builds from a description. Phone app (WB-C4): I found none.

• Image replacement and the phone-only change were not run on the cafe because its chats were gone. The video embed used a placeholder the AI could not verify.

Lighthouse 13.5 with mobile settings ran three times on the home page and one inner page of each site, on the preview address (18 runs), and three times on the desktop setting (18 runs). The median score of the 18 mobile runs is the grade. The pages are the first drafts: the cafe home and menu, the photographer home and pricing, and the shop home and shop pages.

The median of the 18 mobile runs was 89. By page: cafe home 72, cafe menu 93, photographer home 85, photographer pricing 98, shop home 93, shop page 71. The slowest, the shop page, showed its main picture after 7.1 seconds because it loads about 1.3 MB of photos.

The desktop median was 99. Layout shifts were small except on the photographer's home page (0.26). Total blocking time was 0 on every run.

What the pages weigh (WB-P3): 0.3 to 2.3 MB a page. The heaviest files are one 640 KB font file and photos of 200 to 550 KB. All photos are AVIF, there are no render-blocking requests, and JavaScript is 140 to 175 KB.

• Measured on the free preview address, which answers with a noindex header. A published site may be served differently, and the speed from three regions (WB-P1) was not run.

Every plan was read on the in-app plan screen and on the pricing page. The figure graded is the yearly cost of the cheapest plan that connects a custom domain and holds all the pages in my briefs.

The free plan allows 3 websites, 10 free chats per website and no publishing. A fourth website is refused with 'You have reached the maximum number of free websites allowed (3)'. Chats run out with 'Chat Limit Reached: You've used all 10 free chats'.

Basic is $10 a month billed yearly (120 dollars a year; $15 a month billed monthly) for a one-page site. Pro is $20 a month billed yearly (240 dollars a year; $30 a month billed monthly) for as many pages as you need, a shop, invoices and memberships. Each plan covers one website only, and a Pro shop pays a 1% fee on top of Stripe's.

All of my briefs need several pages, so the price I graded is Pro at 240 dollars a year. A domain is 1 dollar for the first year with any plan, then standard prices. The terms say the plan renews at the then-current price. Refunds: a 30-day money-back guarantee.

AI usage (WB-A6): a first build cost no chat; the pages message cost 1 to 2; each later message cost 1 to 2 of the 10 free chats. The paid plans say 'plenty of AI edits' on the pricing page and give 100 (Basic) and 300 (Pro) monthly AI edits on the help page.

• I paid nothing. First-year cost with a mailbox, renewal amounts and itemised bills (WB-M2) need a purchase.