Webflow AI integration that connects through Site Settings

Integration is a word that carries more weight in the Webflow ecosystem than it does on most platforms. Webflow does not have a conventional plugin architecture. There is no wp-content directory to drop files into, no app store where you install an extension and grant it database access, no server-side hook system where third-party code injects itself into page rendering. What Webflow offers instead is a well-defined surface for external scripts: the Custom Code fields in Site Settings. Every third-party tool that wants to be present on a Webflow site — analytics, consent management, retargeting pixels, tag containers, and now AI chat — arrives through those same text areas. The Asyntai Webflow AI integration is designed to land squarely in that surface, behave like a responsible tenant, and deliver value without pretending to be something deeper than what Webflow's architecture allows.

That architectural honesty is worth pausing on, because a lot of tools that claim "Webflow integration" stretch the term. They might require you to embed an iframe inside a Webflow component, or ask you to link a Zapier bridge between your CMS and their service, or insist on a Marketplace app that requests workspace permissions just to render a widget. The Asyntai approach is simpler and deliberately constrained. You paste an async script tag into Head Code or Footer Code under Site Settings, you publish, and the AI loads on every published URL without touching your Webflow project files, your CMS data, or your hosting configuration. The snippet never writes back to Webflow or phones home through your account credentials — it exists purely as a client-side layer that reads what visitors already see and adds a conversational interface on top.

The first thing the integration does after you connect your Webflow URL in the Asyntai dashboard is crawl the published site. This is where the AI stops being generic and starts being useful for your specific Webflow project. The crawler follows every internal link it finds, which on a typical Webflow site means it discovers the homepage, the about page, the services or solutions section, the pricing page, the blog index and every individual blog post rendered from your CMS template, any portfolio or case-study collection, team member pages, FAQ content, contact information, and whatever policy or legal pages are linked from the footer. For Webflow Ecommerce stores, product pages and category listings join the same index. The result is a knowledge base that mirrors the full public footprint of your Webflow project — not a single-page summary, not a homepage-only skim, but the complete published output.

Private material fills in what the public crawl cannot reach. Every Webflow site has institutional knowledge that never makes it to a published page — internal rate cards, discovery-call scripts that shape how the team qualifies prospects, retainer-scope language from past proposals, onboarding checklists that new clients never see publicly, brand-voice guidelines that inform how copy is written but aren't themselves published. You upload those as PDFs or paste them as plain text directly into the Asyntai knowledge base. The AI treats them identically to the crawled pages: as source material it references when a conversation calls for it, without ever exposing the documents to your Webflow visitors. That blend of crawled pages and uploaded materials is what gives the AI the depth of a colleague who has been through the company onboarding, not a cold reader who scanned the marketing copy once.

Behavioral rules are the third input that shapes how the AI operates within your Webflow integration. These are plain-English instructions you write once and apply across all conversations. "When a visitor asks about project timelines, always check whether they have a fixed launch date before suggesting a scope." "Route any question about annual retainers to the contact form on the retainers page." "If someone mentions a competitor by name, acknowledge it neutrally and pivot to our differentiators on the comparison page." "Always use first-person plural." The AI applies these rules consistently regardless of visitor language, session length, or question complexity. For agencies managing Webflow sites on behalf of clients, these rules are effectively a lightweight brand-voice layer that keeps the AI aligned with the client's positioning without requiring the agency to micromanage every conversation.

The integration surface itself — Head Code versus Footer Code — is a choice worth making intentionally rather than defaulting. Head Code means the AI script begins downloading before the page body renders. The async attribute ensures it does not block rendering, but the download starts early. Footer Code means the script begins downloading after the body has parsed, which puts hero images, Lottie animations, and above-the-fold content further ahead in the priority queue. For design-heavy Webflow sites where the initial visual impact matters — studios, fashion labels, real estate showcase sites — Footer Code is usually the better slot. For content-driven sites where the chat is a core conversion tool — lead-gen landing pages, SaaS marketing microsites, services pages — Head Code gets the widget ready fractionally sooner. The absolute difference is small, but having the option matters to teams who care about render priority.

Per-page versus site-wide integration is another axis of control. Webflow offers Custom Code at two levels: the site-wide fields in Site Settings (which propagate to every published URL), and per-page code fields accessible from the page settings panel in the Designer. You can run the Asyntai snippet site-wide for blanket coverage, or restrict it to specific pages by pasting it only in those page settings. Agencies often start with a handful of high-traffic pages — the services landing page, one or two flagship case studies, the pricing page — to validate the integration before rolling it out project-wide. Solo operators running a single portfolio usually paste site-wide from day one because every page is a conversion surface and full coverage costs nothing at the free-tier message volume.

Language handling is woven into the integration rather than layered on top. The AI widget ships with interface strings localized across 36 languages, and the model picks up on the written language of the opening message and responds in kind. A Munich-based prospect who writes in German sees German throughout the widget and the conversation. A Sao Paulo visitor typing Portuguese gets Portuguese. The underlying model translates between your English Webflow content and the visitor's native tongue at inference time — no Webflow Localization project, no per-locale snippet, no third-party translation service in the middle. For Webflow projects that pull organic traffic across borders (common, given that the platform outputs fast, crawlable markup that ranks broadly), this single-snippet multilingual behavior fills a gap most site owners never even track.

Lead capture through the AI integration works differently from a traditional Webflow form block. A form block on a Webflow page is a deliberate action the visitor has to seek out, navigate to, and complete with structured fields — many visitors with genuine questions never reach it. The AI captures contact details mid-conversation, at the moment when the visitor's engagement is highest and their objections have already been addressed. The captured lead includes the full transcript — every question they asked, every answer the AI provided, every signal about their intent. That transcript context transforms follow-up from cold outreach to warm continuation. Leads land in the Asyntai dashboard and, with email notifications enabled, directly in the inbox of whoever handles sales or client relationships for the Webflow project.

User Context integration is the mechanism for Webflow projects that handle authenticated visitors. Memberstack-powered membership sites, SaaS applications using Webflow for their marketing shell, and gated community platforms all have visitors who are logged in when they encounter the AI. On Standard and Pro plans, you populate a window.Asyntai.userContext object in your own Webflow Custom Code before the widget initializes. That object can carry name, email, plan tier, account creation date, renewal status, or any custom attribute your membership layer already knows. The AI weaves this context into its replies: a paid member asking about an upgrade sees a reply that references their current tier; a trial user asking about features sees a reply scoped to what the trial includes. Asyntai reads only what your code passes; nothing is pulled from Webflow or the membership provider without your involvement.

Conversation analytics close the feedback loop. Every AI dialogue on your Webflow site is logged, timestamped, and clustered by topic and language in the Asyntai dashboard. Within weeks, patterns emerge that no analytics tool would surface on its own — repeated pricing questions flag a page gap, geographic inquiries reveal missing geographic statements on the homepage, and unexpectedly active CMS posts identify conversion assets you did not know you had. These patterns come from actual visitor phrasing, not pageview proxies, which makes them far more actionable for Webflow teams iterating on copy weekly.

Multi-site management mirrors how agencies actually run Webflow books of work. Free covers one project. Starter handles two at $39 per month with 2,500 messages. Standard extends to three with User Context. Pro scales to ten independently trained integrations. Each project keeps its own isolated knowledge base, preventing content crossover between unrelated clients. Warning emails arrive before the monthly message ceiling so a traffic spike on one project does not silently disable the integration across the rest.

The path to a working Webflow AI integration fits inside a single working session. Register a free Asyntai account, copy the snippet, open the Webflow Designer, navigate to Site Settings, paste the snippet into Head or Footer Code, save, publish. Return to Asyntai, submit the Webflow project URL so the crawler indexes every published page. Upload any private documents you want the AI to reference. Draft a handful of tone and escalation rules in everyday language. Adjust the widget color, position, and greeting to match the Webflow project's design language. Run a handful of test conversations from the live URL to confirm voice and accuracy. Switch on lead capture notifications. The whole sequence is complete before lunch, and from that point forward the Webflow AI integration is handling visitor questions, converting hesitant browsers into qualified leads, and generating the kind of conversation data that makes every future edit to the Webflow project more informed than the last.