# How to launch and test a campaign landing page without waiting for development > From the idea to the result in five steps: decide what the page must achieve, build it by chatting with the AI, publish it on your domain, A/B test it with a workflow and read the verdict without fooling yourself. - Source: https://bixlow.com/en/blog/launch-and-test-a-campaign-landing-page-without-waiting-for-development - Language: en - Published: 2026-11-08T10:00:00.000Z - Updated: 2026-11-09T14:38:47.213Z - Author: fernando - Topics: landing pages, campaign landing page, A/B testing, CRO, AI landing page builder - Publisher: Bixlow --- To launch a campaign landing page and test it without waiting on engineering you need four things: to know what should happen on the page, a way to build it yourself, somewhere to publish it under your own domain, and a way to measure it and compare versions with the same goals as the rest of your site. In Bixlow all four live in the same project: you ask for the page in a chat, publish it on your domain and, since the Bixlow script is built in, a workflow can run an A/B test on it. Below is the full route, step by step, with what's worth deciding at each one. ## Step 1: decide what should happen on the page Before anyone writes a line, note down three things. It sounds obvious, but it's what gets skipped when the campaign launches on Monday: - **Who it's for.** "People arriving from the winter sale ad on Instagram" is not the same as "any visitor". The more specific, the easier the headline is to write. - **What the ad promises.** The landing page has to repeat that promise at the top, in the same words. If the ad says "up to 40% off coats", the page can't open with the brand's story. - **What counts as success.** A click on "See the collection", a submitted form, a visit to the thank-you page. That will be your [goal](/en/platform/goals), and it pays to settle it now because it's what will decide the test in step 4. With that written down, the rest goes fast. ## Step 2: ask for the page in the chat In the project's landing pages section you create a new one and tell the AI what you just wrote down. For example: "Landing page for the winter sale campaign. It's for people coming from the Instagram ad. At the top, the ad's offer; then three featured products and the shipping terms; at the end, a button to the collection." The AI writes the page and you watch it take shape in the preview. From there you adjust it in three ways: - **By pointing.** Turn on the picker, click the element you want to change and ask for the change: "this headline, shorter", "this button, below the price". The AI changes that element and nothing else. - **By touching the code.** The HTML, CSS and JS sit in separate tabs. If you know what you want, edit it yourself and the draft saves on its own. - **By attaching images.** Campaign photos you attach in the chat are saved to the project's assets and the AI uses them on the page. Before moving on, look at it at tablet and mobile size from the preview itself. Most traffic from a social ad arrives on a phone, and that's where a long headline or a buried button costs the most. Finally, open the settings: search title and description, language, social sharing image and whether it should be indexed. For a campaign page that only lives a few weeks, leaving it out of the index is often the sensible choice. ## Step 3: publish it on your domain Publishing creates a version of the page that no longer changes and puts it live in one step: whoever visits gets the previous one or the new one, never a mix of both. If you spot a mistake the next day, you go back to the previous version in one click. When the campaign ends, you unpublish it and the versions stay saved for next time. As soon as it's published, the landing page answers at a Bixlow address. For a campaign it's usually better to have it under your own domain, since that's what visitors see in the address bar and in the ad. There are two ways: - **A path on your domain**, like `yourbrand.com/sale`. Bixlow gives you the exact snippet for a Cloudflare Worker, Nginx or Apache, whichever your site uses. Only that path is touched; the rest of the site stays as it is. - **A subdomain**, like `sale.yourbrand.com`, with a CNAME record and a TXT record that proves the domain is yours. The certificate is issued once that record is verified. Either way, Bixlow checks the page really is served where you say before marking it as verified. If someone else runs your server, the snippet is what you send them: it's short and needs nothing else changed. ## Step 4: test it with a workflow This is the difference from a standalone page builder. The landing page has `bixlow.js` embedded, the same script that runs on your site, so to Bixlow it's one more page of the project: your goals count its conversions, your audiences apply and workflows run on it. To compare two versions of something, say the headline, you create a workflow whose URL rule points at the landing page and switch it to A/B test mode. The current version is the Original; in the variant you ask the chat for the change: "in variant B, change the headline to one about free shipping". Pick the goal from step 1, the split and, if you like, an audience, and publish it. A few things worth knowing: - **The split happens in the browser.** Each visitor lands in one version and keeps seeing it on later visits. - **Test one thing at a time.** If you change the headline, image and button together and the variant wins, you won't know which of the three did it. - **Test what visitors notice.** The offer, the headline, the order of the sections or the button text move more than a shade of colour. If you'd rather not build the workflow by hand, you can ask for it in your own words in the project chat and the AI builds it for you as a draft. Publishing is still your call. ## Step 5: read the result without fooling yourself While the test runs, the report shows how many visitors each version has seen and how many converted. What it won't do is name a winner the moment one version pulls ahead. The verdict comes from [Bayesian statistics](/en/platform/statistics) and only appears once three conditions are met: a minimum sample per version, a minimum number of days (so a weekday-only sample can't decide) and a high probability of being best with a small expected loss. It also checks that the real split matches the one you asked for; if it doesn't, it warns you before you trust the result. Next to each version you'll see its probability of being best, the relative change against the Original and a credible interval for that change. If the interval runs from -3% to +18%, the variant looks promising but may still improve nothing: wait. When there's a winner, you declare it and it becomes the version everyone sees. And if there's no winner, that's an answer too: the new headline changes nothing, so the next test goes after something else. ## Mistakes we see often 1. **Sending the ad to the home page "in the meantime".** A two-week campaign that spends its first week pointing at the home page has spent half its budget on the wrong page. With the landing page ready in a day, there's no "meantime". 2. **A landing page that doesn't repeat the ad.** If visitors can't find what they were promised within three seconds, they leave. Copy the ad's headline, word for word, into the first version. 3. **Publishing without a goal.** Without a defined goal, the landing page is a nice page with no way of knowing whether it works. Define it before you publish. 4. **Checking the test after two days and stopping it.** The first days almost always mislead. Let the minimum sample and duration pass. 5. **Not checking it on mobile before publishing.** That's what the preview sizes are for. ## Frequently asked questions **Do I need to code to build a landing page in Bixlow?** No. The AI writes the page and changes it when you ask, by pointing at the element in the preview. If you know HTML and CSS, you can edit the code directly. **Does the landing page replace my site?** No. It lives next to it: at a Bixlow address, on a path of your domain or on a subdomain. The rest of your site doesn't change. **Is it good for SEO?** The page ships with its content in the HTML, a title, description, canonical and a project sitemap, so it's technically ready for a search engine. Whether it gets indexed and where it ranks is up to the search engine. For short-lived campaign pages, consider leaving them out of the index. **Can I use the audiences and goals I already have?** Yes. Since the landing page includes `bixlow.js`, the project's goals and audiences apply just as on any other page of your site. **Which Bixlow plan do I need?** Landing pages are in every paid plan, from Standard up. The free Basic plan doesn't include them. The details are on the [landing pages](/en/platform/landing-pages) page. --- ## Meet our AI CRO agent What you just read, put to work by an agent: it understands your visitors, decides what to test or personalize, builds it on your site without a developer and tells you when a result is real. https://bixlow.com/en/ai-cro-agent