VWO AI review (Oct 2026): AI copilot for A/B testing and website experiments

What this VWO page covers

This page is for a marketer, a conversion rate optimization specialist (someone whose job is getting more visitors to sign up or buy), or a product manager who runs website tests in VWO — or is weighing it up after the rename. The question it answers: which Wingz features save time between “something looks wrong on this page” and a properly measured test, what each one costs in AI credits, and where a person still decides what gets tested and when.

On naming: VWO and AB Tasty announced they were joining forces on 20 January 2026, and launched the shared Wingify brand and platform on 16 September 2026. Wingify’s product pages now say “Wingify Experimentation” instead of VWO; its help center still mentions “Copilot” in places, and its page scanner still identifies itself as “VWO.” Existing VWO contracts and pricing were said to be unchanged by the merger. Wingz itself comes in two tiers: Wingz Core, included with every paid Wingify product, and Wingz Advanced, a paid upgrade.

Everything below comes from Wingify’s Wingz and pricing pages, its January and September 2026 announcements, and its help-center articles (most updated 17–25 September 2026), as we read them on 9 October 2026. Wingify does not publish prices. Figures such as “70–90 billion events a day” and “4,000+ customers” are Wingify’s own claims.

Major features (with how someone uses them)

Examples at a glance: each feature has a concrete example from a real job plus a short Try this you can run this week.

1. Three hypotheses from one heatmap, each with a metric and sample size (primary)

What it is: A heatmap colors a page by where visitors click, move, and scroll. While viewing one, you click Wingz, allow a screen capture, and in about a minute it returns a Heatmap Analysis Report: insights sorted by High, Medium, and Low impact, each with an observation and a suggested action, plus links to the part of the page it means. Wingz guesses a business goal from the page; you can type your own. Any insight can be saved as a hypothesis — a testable statement such as “moving the price above the fold will raise trial sign-ups.” The page needs at least 100 clicks, and Wingify says the feature is in early access and rolling out in phases.

Before building any test, write two things for each hypothesis. The metric: the one number that decides the winner (trial sign-ups, not “engagement”). The sample size: how many visitors each version needs before the result means anything — work it out with a free sample size calculator from your page’s current conversion rate and the smallest improvement worth acting on. If the page can’t reach that number in a few weeks, pick a busier page or a bigger change. Wingify’s credit guide puts a single-page heatmap analysis at about 200 credits.

Concrete example (growth marketer on a pricing page) A growth marketer runs Wingz on the pricing page heatmap. The report’s top insight says visitors click the plan comparison table but rarely scroll to the sign-up button below it. She saves three hypotheses, then writes for each: metric “trial starts,” current rate 3.1%, smallest lift worth it 15%. The calculator says about 23,000 visitors per version, or 46,000 in total; the page gets about 60,000 a month, so one test fits in roughly three weeks. She keeps the strongest hypothesis and parks two for later.

Try this

Primary workflow:

Ask VWO’s copilot for three hypotheses from one heatmap. Write the metric and sample size for each before you build it. Drop any hypothesis whose page can’t reach that sample size within four weeks.

3. AI-built variations from a prompt, a design file, or a click

What it is: Wingz can build a test variation from a sentence (“add a sticky countdown banner”), from a Figma design file or screenshot, or by clicking an element on the page and saying what to change. In the Visual Editor, once you have edited one variation, Remix (the help article still calls it “Remix Your Variation with Wingify Copilot”) can generate more versions — either automatically from your first edits or guided by your own prompt — which you preview and add if you like them. Credit estimates in Wingify’s guide: about 50 for a headline change, 100 for adding a pop-up, 150 for a large section from a Figma file.

Wingify’s own advice is to test one hypothesis at a time and start with small changes so you can tell what made the difference. Three AI headlines in one test split your traffic three ways and need more visitors; two good versions are usually better than five quick ones. Preview every variation on mobile, where AI layout changes break most often.

Concrete example (SaaS marketer, trial landing page) A marketer rewrites the landing page headline by hand as variation one, then asks Remix for two more using the prompt “make the benefit concrete, under eight words.” She discards one that promises a feature the product doesn’t have, keeps one, and runs a three-way test only because the page has enough traffic for it.

Try this

One by hand, two by AI:

Make one variation by hand, then ask Remix for two more with a specific prompt. Preview all three on mobile and keep at most two that make claims your product can back up.

4. Wingz Chat, agents, and how much autonomy to give

What it is: Wingz Chat (included with paid plans) answers questions about your tests in plain language — “compare desktop vs mobile in this campaign” — reads attached screenshots or briefs, and can audit a draft campaign: its traffic split, hypothesis, and main and secondary metrics. Agents are saved specialists (Wingify lists Insight, Hypothesis, Experiment, Guardrail, and more, or build your own), and Workflows chain agents on a schedule or trigger. Wingify describes autonomy as a dial: Propose and approve, Act inside guardrails, or Autopilot, with an approval gate, guardrail metrics, traffic caps, a kill switch, and a permanent holdout group always on.

A holdout is a slice of visitors who never see changes, so you can measure the total effect. Start every team at “Propose and approve,” where a person launches each test, and write down who that person is. Recurring workflows charge credits every time they run — Wingify’s example is about 70–100 credits for a weekly heatmap summary — and AI Workflow and AI Autopilot are Wingz Advanced features on the pricing page.

Concrete example (experimentation lead before a launch) Before launching a checkout test, an experimentation lead asks Wingz Chat to audit the draft. It flags that the secondary metric is set to page views instead of completed orders and that mobile visitors are excluded. She fixes both, then launches it herself; the team’s autonomy setting stays at Propose and approve.

Try this

Audit before launch:

Ask Wingz Chat to audit one draft campaign before launch — goal, traffic split, and metrics. Fix what it flags, and name the one person who presses launch.

5. Synthetic A/B tests and the Wingify MCP

What it is: Synthetic A/B testing (Wingz Advanced only) lets you upload screenshots of two page versions and have AI-generated personas — “Price-Sensitive Shopper,” “Growth Marketer,” or your own — answer evaluation questions about them. Wingify calls it “directional validation” that “does not replace live user traffic testing.” The Wingify MCP server (Model Context Protocol, a standard way to connect AI assistants to other software) lets Claude, ChatGPT, Cursor, or automation tools such as n8n and Zapier read your campaigns, heatmaps, and recordings, signing in with your Wingify account (OAuth) rather than copying API keys. Behavior-analytics data through MCP needs Wingz Advanced.

Use synthetic tests to discard weak ideas, never to declare a winner. Blur anything confidential in screenshots first; Wingify says it doesn’t detect sensitive content in uploads. Through MCP, Wingify charges credits only for fetching data (for example, about 45 credits to analyze a running campaign), but an MCP client can also create, start, or pause a test, so decide who may connect and for what.

Concrete example (conversion specialist, weekly review) A conversion specialist connects the Wingify MCP in Claude and asks which campaigns in the last three months had no clear winner. It lists seven. She opens each in Wingify, confirms four simply lacked traffic, and closes them herself; she keeps start and pause actions inside Wingify.

Try this

Read-only first:

Connect the Wingify MCP in Claude or ChatGPT and ask which campaigns had no winner in the last three months. Check each in Wingify, and agree as a team that tests are started and stopped only inside Wingify.

6. Plans, credits, and where VWO fits

What it is: Wingify’s pricing page lists Web Experimentation in Growth, Pro, and Enterprise plans with “Schedule a demo” and no prices, and says pricing is custom, based on monthly active users (behavior analytics uses tracked users, a subset). Every paid plan includes Wingz Core with 2,500 AI credits a month that don’t roll over; Wingz Advanced is a paid upgrade with 8,000 or 15,000 credits a month, optional 5,000-credit packs, and an Enterprise Data Control add-on (run Wingz on your own model, or keep AI processing in the US or EU). Advanced and add-ons are switched on through your account manager, not self-serve.

VWO fits marketing and conversion teams who test website pages and want heatmaps, recordings, and testing in one place with AI help. Teams whose tests are owned by engineering and run inside the product often compare Statsig or LaunchDarkly; enterprise marketing teams compare Optimizely; teams that mainly need behavior insight look at Contentsquare or Fullstory.

Concrete example (boundary) A team plans four heatmap analyses (about 800 credits) and two AI-built campaigns (about 1,100) a month, roughly 1,900 credits, inside Core’s 2,500. A team that wants weekly workflows across twenty campaigns plus synthetic tests will need Advanced, so it asks for that price up front.

Try this

Credit budget:

Estimate a month of Wingz use with Wingify’s credit guide — about 200 per heatmap analysis and 550 per AI-built campaign — and compare it with Core’s 2,500 credits before you ask about Advanced.

Pricing context (October 2026)

Wingify (formerly VWO) does not publish prices. Its pricing page on 9 October 2026 said “No fixed plans. Just a custom proposal,” based on monthly active users, products, plans, and setup scope. Web Experimentation & Rollout shows Growth, Pro, and Enterprise with feature lists (Growth includes the visual and code editors, unlimited experiments, and Wingz Core; Pro adds guardrail metrics and multivariate tests, which test several changes at once; Enterprise adds approval workflow and single sign-on). There is a 30-day free trial with no card.

AI: Wingz Core is included with every paid product (Wingz Chat analysis and MCP without behavior-analytics data; 2,500 credits a month per Wingify’s help center). Wingz Advanced is paid (MCP with behavior data, AI Workflow, AI Autopilot, Synthetic A/B tests; 8,000 or 15,000 credits a month), with credit packs and Enterprise Data Control as add-ons. Ask for a quote that shows your monthly active users, the products you need, and whether you need Advanced.

Uncertainty: Wingify lists no prices, so any figure you see elsewhere is not from Wingify or VWO. The rebrand is recent (16 September 2026) and product names, help articles, and packaging may keep changing; heatmap insights are in early access. Credit costs are Wingify’s own estimates and vary with each request. We read everything on 9 October 2026.

When it’s useful (honest take)

Useful when a marketing or conversion team runs website tests every month, has enough traffic to reach a sample size in a few weeks, and wants AI to speed up the slow parts — reading heatmaps, drafting ideas and variations, checking draft campaigns — with a named person approving and launching every test.

Less useful when your pages get too little traffic for tests to finish, your experiments live in product code owned by engineers, or nobody tracks credit use. Synthetic tests are a filter, not evidence; teams that treat them as results will ship wrong winners.

Related options (educational)

FAQ

Is VWO now called Wingify?

Yes, in effect. VWO and AB Tasty announced their merger on 20 January 2026 and launched the shared Wingify brand and platform on 16 September 2026. vwo.com now forwards to wingify.com, and products are named Wingify Experimentation and so on. Wingify said existing VWO contracts and pricing were unchanged by the merger.

What is VWO’s AI copilot called now?

Wingz. It analyzes heatmaps, recordings, and campaigns, suggests test ideas, builds variations from prompts or design files, audits draft campaigns in Wingz Chat, and runs agents and workflows. Some Wingify help articles still use the older Copilot name.

How much does VWO cost in October 2026?

Wingify does not publish prices; it offers custom proposals based on monthly active users and products, with a 30-day free trial. Wingz Core (2,500 AI credits a month) is included with paid plans; Wingz Advanced is a paid upgrade. See wingify.com/pricing.

Will Wingz change my website by itself?

Wingify says no: Wingz handles analysis and suggestions, your team decides what gets tested and when, and nothing goes live without explicit approval. Higher autonomy levels exist (Act inside guardrails, Autopilot), so decide your team’s setting and name who launches tests.

VWO or Optimizely?

Both run website experiments with AI help. VWO (Wingify) bundles heatmaps, recordings, and testing with Wingz credits included on paid plans; Optimizely focuses on large marketing and content programs with its own AI agents. See Optimizely Opal.

Optional next step (if seats overlap)

If you pay for a testing tool, a separate heatmap and recording tool, a survey tool, and a general AI assistant that writes the same test ideas, that overlap is worth a look. Map the overlap with the $19 AI toolkit audit or the $9 AI toolkit spreadsheet. Email Audit $19 or Toolkit spreadsheet $9.

Checkout is live via Stripe. After you pay, we email delivery (spreadsheet) or intake next steps (audit).

Related: Whatfix AI · WalkMe AI · Optimizely Opal · Statsig · LaunchDarkly AI · Contentsquare AI · Fullstory AI · PostHog AI · Claude · ChatGPT · Stack map for product ops · How-to: Turn VWO heatmap insights into a test with Wingz, and have the CRO lead approve the hypothesis and variation · ops hub