Google Pomelli and Stitch vs Webflow: How AI Website Builders Just Reshaped the Agency Conversation

Google announced Pomelli at I/O 2026 today. The pitch is that AI agents can design your brand book and launch a website, end to end. Stitch got a meaningful update too, with real-time layout reflow that closes some of the gap with Framer Workshop and Webflow AI. Both products are aimed squarely at the early-stage founder who is currently choosing between hiring a Webflow agency and just having AI build it.

I have spent the evening kicking the tires on both. Here is the honest take from a working Webflow partner. Some of this is genuinely impressive. Some of it is the same launch-day overpromise we have seen a dozen times. Here is the breakdown.

What did Google announce about Pomelli at I/O 2026?

Quick answer: Per the official Google I/O 2026 announcements page, Pomelli launched as a product that introduces AI agents which can help you design your brand book and launch a website. The system is powered by Gemini 3.5 Flash with deeper Antigravity 2.0 integration for the generation pipeline. The brand book output covers logo direction, colour palettes, typography pairings, and tone-of-voice samples.

The website-launch flow takes the brand book and generates a landing page or a small marketing site. The output is hosted on Google infrastructure. There is no Webflow CMS underneath, no Framer-style component library, and no native way to migrate the site to another platform if you outgrow it. That is the most important detail and the one that did not appear in the keynote slide.

How does Google Stitch's new real-time reflow compare to Framer Workshop and Webflow AI?

Quick answer: Stitch's real-time layout reflow watches you make changes and updates the underlying layout structure as you work, similar to how Framer Workshop handles fluid layouts. The difference is that Stitch is generative-first. You describe the change and the layout responds. Framer Workshop still expects you to manually arrange components with snapping helpers. Webflow AI sits between them, offering generation for sections but expecting you to drive the Designer for layout decisions.

For raw speed of getting a single landing page live, Stitch with the new reflow is genuinely fast. For building a B2B SaaS marketing site with five pages, a blog, a customer story collection, and proper SEO structure, Webflow's Designer plus the May 13 Audit panel still beats both Stitch and Pomelli by a wide margin.

Will Pomelli actually replace a 20,000 dollar Webflow agency project?

Quick answer: Not yet. Pomelli can replace the 2,000 dollar landing page commission and the 5,000 dollar logo plus single-page-site package. It cannot replace a 20,000 dollar engagement that includes positioning work, content strategy, a multi-page CMS-driven marketing site, conversion-rate-optimised forms, ongoing analytics, and the human judgment about what to build versus what to cut. The gap is judgment, not generation.

I am not worried about the high end of the agency market for the same reason senior architects are not worried about AutoCAD. The tool generates the artefact. The judgement decides which artefact to generate. Pomelli does the first part well and the second part not at all.

Where does Pomelli fail today for B2B SaaS marketing sites?

Quick answer: Pomelli fails on five things that matter for B2B SaaS marketing sites. It does not handle CMS-driven blog content at scale. It does not produce site-architecture decisions that anticipate growth. It does not integrate with the analytics, CRM, and marketing automation tools that B2B teams actually use. It does not produce hand-off-able code if you want to migrate. And it does not understand the buyer journey beyond the visible page.

These are not minor gaps. They are the entire reason a B2B SaaS company hires a Webflow agency in the first place. A landing page is not a marketing site. A brand book is not a positioning strategy. Pomelli ships the artefacts. The strategy underneath them is what founders are actually paying agencies to deliver.

How does Pomelli's brand book agent compare to a Figma plus Webflow workflow?

Quick answer: Pomelli's brand book agent produces a polished-looking deliverable in roughly fifteen minutes. A Figma plus Webflow workflow with a designer typically takes two to four weeks and costs between 5,000 and 15,000 dollars. The Pomelli output is competent but generic. The human-designed output reflects the actual positioning of the company and ties into the marketing site through shared design tokens. Both have a place. They are not direct substitutes.

I would use Pomelli to draft a starting point for a brand exploration. I would not ship a Pomelli output as the final brand book for a B2B SaaS company with a real go-to-market plan. The conversation between designer and founder that produces a brand book is doing more than generating colour palettes. It is forcing the founder to articulate what the company is.

What does Webflow's Audit panel still do that Pomelli does not?

Quick answer: The Webflow Audit panel that shipped on May 13, 2026 runs SEO, performance, accessibility, and AEO checks against your live site and surfaces specific issues to fix. Pomelli generates the site but does not audit it post-launch. Webflow customers who adopted the Audit tool saw 75 percent more monthly organic traffic per Webflow's own reporting. Pomelli has no equivalent ongoing-quality layer.

This is the most underrated detail. Building a site is one job. Maintaining it is a much longer one. The Audit panel is the difference between a Webflow site that compounds value over years and a Pomelli-generated site that is impressive at launch and slowly decays because nothing surfaces the problems.

How should agencies reprice discovery and brand work after Pomelli?

Quick answer: The repricing I am making to my own engagement structure is to compress the discovery phase from two weeks to one week using AI-assisted research, drop the brand book deliverable as a separate line item, and reinvest the saved time in deeper positioning work and longer post-launch optimisation. The hourly rate stays the same. The deliverables shift. The total project value stays roughly equivalent.

For agencies that were charging 5,000 dollars for a brand book, that line item is going to compress. The work it represents is still valuable, but the deliverable on its own no longer commands the same price now that Pomelli can produce a draft for free. Agencies that survive this either move up-market into deeper strategy work, or down-market into volume.

Is this Google's WordPress-killer or another developer toy?

Quick answer: Neither. Pomelli is not a WordPress-killer because WordPress's strength is the plugin ecosystem and the publisher workflow, neither of which Pomelli addresses. It is not a developer toy because the target user is explicitly the non-technical founder. Pomelli is best understood as a Squarespace competitor with better generation tooling, targeted at people who would otherwise use Squarespace or Wix.

The interesting market position is that Pomelli probably takes share from the bottom end of Wix and Squarespace more than from Webflow or Framer. The founder who would have built on Squarespace in 2023 now has Pomelli as an option. The founder who hired a Webflow agency in 2023 still has the same reasons to do so.

How does Pomelli sit alongside Webflow MCP and Claude Code?

Quick answer: Pomelli and the Webflow MCP plus Claude Code workflow target completely different users. Pomelli is for the non-technical founder who wants a website without hiring anyone. Webflow MCP plus Claude Code is for the technical operator who is already running a Webflow site and wants to automate content production, audits, and updates at scale. The two stacks do not compete.

I am running Webflow MCP plus Claude Code daily. I am also experimenting with Pomelli for quick landing page tests. The mental model I use is that Pomelli is for the first website, while Webflow MCP plus Claude Code is for the hundredth piece of content. Both have a place. Confusing them produces bad recommendations.

What should B2B founders evaluate over the next 90 days?

Quick answer: Three evaluations. First, run Pomelli on a throwaway domain and see what it produces, just to calibrate your sense of what AI website builders can and cannot do. Second, audit your current Webflow site through the May 13 Audit panel. Third, talk to your Webflow agency about how their service offering is changing post-I/O 2026.

The founders who handle this well are the ones who treat Pomelli as a calibration tool, not a threat. The calibration matters. If your current marketing site cannot beat a Pomelli output, your marketing site has a positioning problem, not a tooling problem. Fix the positioning. The tool will follow. My Webflow Conf 2026 preview covers what the Webflow side of this story looks like, and my piece on client handoff process covers how agencies are restructuring their deliverables.

The honest take after a day of looking at Pomelli, Stitch, and the new Antigravity 2.0 integrations is that Google built impressive generation tools that do not yet replace the judgement work that justifies hiring a Webflow agency. The artefacts get faster to produce. The strategy underneath them still has to come from somewhere. If you want to walk through what your own Webflow site versus AI-generated alternatives looks like, let's chat.