Implementation overview
Create LLM Content Feeds in Webflow: /ai-feed.json Guide (2026)
An LLM content feed is a machine-readable file — typically at a URL like /ai-feed.json or /llms-feed.json — that gives AI systems structured, up-to-date access to your site’s content. Where an llms.txt file declares your policies and sitemap reference, a content feed delivers the actual data: titles, summaries, publication dates, authors, and URLs for every piece of content on your site.
This matters because AI systems that retrieve content in real time — like Perplexity’s online mode or ChatGPT with Browse — benefit from a clean, structured endpoint far more than from crawling individual HTML pages. A JSON feed gives them precisely formatted data they can parse without needing to strip navigation, ads, and boilerplate from your page HTML.
In Webflow, content feeds can be built as dynamic collection pages that output JSON instead of HTML, automatically staying in sync with your CMS as you publish, update, or archive content. No external infrastructure required.
How to do it on Webflow?
1. Create a dynamic JSON feed page in Webflow
Webflow doesn’t natively output JSON, but you can approximate a structured feed using an Embed element on a collection-bound page with a custom URL structure:
• Create a new static page with the slug ai-feed (accessible at yoursite.com/ai-feed)
• On this page, add a Collection List bound to your primary content collection (e.g. Blog Posts)
• Inside each collection item, use an Embed element to output structured data per item:
“title”: [CMS title field],
“url”: yoursite.com/[slug],
“excerpt”: [CMS summary field],
“authori”: [author name],
“published”: [published date],
“updated”: [last updated date],
“category”: [CMS category field]
This produces a human-readable structured page rather than true JSON, but it gives AI crawlers a clean, parseable data layer without the noise of navigation and UI elements.
2. Structure the feed data fields
For each content item in your feed, include these fields as a minimum:
• title — the page title or SEO title
• url — the full canonical URL
• excerpt — a 100–200 word summary (not the full body — enough for context, not a full replacement)
• author — author name and role
• published_date — ISO 8601 format
• updated_date — ISO 8601 format
• categories — topic tags or collection categories
• content_type — Guide, Tutorial, Reference, Opinion, etc.
The excerpt is the most important field for AI systems — it’s what gets read and potentially cited. Write excerpts as standalone summaries, not as teasers that require clicking through.
3. Add access controls and rate limiting
A public AI feed can be crawled aggressively. Add basic protections:
• Add your feed URL to your llms.txt file — this tells AI systems where the feed is and how they may use it
• In Webflow’s Project Settings → SEO, ensure the feed page is not blocked by robots.txt — it needs to be crawlable
• Include a License and Contact field at the top of the feed page pointing to your attribution policy
• Consider adding a last-modified header via Webflow custom code so AI systems know when to re-fetch the feed
4. Expand to multiple feeds by content type
A single feed works for small sites. For larger sites with multiple content collections, create separate feeds per type:
• /blog-feed — blog posts and articles
• /guides-feed — long-form how-to content
• /tools-feed — tool listings or resources
• /ai-feed — a master feed combining key items from all collections
Reference all feed URLs in your llms.txt file so AI crawlers can discover each one without needing to find them via sitemap parsing. Pair this with embedded author IDs and canonical URLs to ensure each feed item has unambiguous ownership metadata.
5. Automate feed updates with the Webflow MCP server
The feed’s value comes from being current. Use the Webflow MCP server to trigger a feed refresh whenever a new CMS item is published or an existing item is significantly updated — ensuring AI systems that cache your feed see fresh data rather than a stale snapshot.
Also add your feed URL to your sitemap and submit to Google Search Console alongside your standard sitemap — search crawlers may use it too. Complement the feed with optimised Open Graph metadata so each URL in the feed also has a clean preview when retrieved individually.
Frequently Asked Questions
What is an LLM content feed?
An LLM content feed is a structured, machine-readable file (typically JSON or a structured HTML page) that provides AI language models with organised access to your site’s content — including titles, summaries, authors, dates, and URLs. It’s designed to give AI retrieval systems a clean data source rather than requiring them to parse your full HTML pages with all their navigation and UI noise.
Is an LLM content feed the same as an RSS feed?
Not exactly, but they serve a similar purpose. RSS feeds are XML-based and designed for feed readers. LLM content feeds are typically JSON-based and optimised for AI retrieval systems, with richer metadata fields (content type, author credentials, update dates) and structured excerpts rather than full content. Many sites start by repurposing their RSS feed and adding AI-specific fields to it.
Will AI systems automatically discover and use my content feed?
Only if you tell them where it is. Reference your feed URL in your llms.txt file and sitemap. Some AI systems (particularly those with real-time web retrieval) will crawl it if they find a reference in your metadata. There is no universal standard yet for AI feed discovery — explicit declaration in llms.txt is currently the most reliable signal.
Should the feed include full article content or just excerpts?
Excerpts — 100 to 200 words per item. Full content in a feed creates two risks: it enables AI systems to use your content without users visiting your site (reducing traffic), and it can trigger duplicate content signals if the feed content exactly mirrors page content. Well-written excerpts are enough context for AI systems to understand and cite the source page.
Sources
• JSON Feed — A format similar to RSS and Atom but in JSON
• llmstxt.org — Specification for AI content discovery files
• Google — Sitemaps overview
Do's
✅ Include full canonical URLs in every feed item: Relative URLs break when the feed is parsed outside your domain context
✅ Write excerpts as standalone summaries: 100–200 words that give an AI system enough context to cite the source without needing to visit the page
✅ Reference your feed in llms.txt: AI crawlers won’t find your feed unless you explicitly declare its location
✅ Include publication and update dates in ISO 8601 format: Machine-readable dates allow AI systems to evaluate content freshness accurately
✅ Keep the feed current with automated updates: A stale feed is worse than no feed — outdated excerpts lead to outdated AI citations
Do's
❌ Don’t include sensitive or private content: Only include publicly accessible content you want AI systems to index and potentially cite
❌ Don’t publish full article bodies in the feed: Excerpts protect your traffic by giving context without replacing the page visit
❌ Don’t block the feed page in robots.txt: A blocked feed is invisible to AI crawlers — verify your robots.txt doesn’t accidentally exclude it
❌ Don’t use complex nested data structures: Keep JSON output flat and parseable — deeply nested objects are harder for AI systems to process reliably
❌ Don’t forget to validate feed items have required fields: Items missing a URL or title break AI parsers — add CMS validation to prevent publishing incomplete feed entries
Tools
