Edge AI Delivery
Serve different content variants to AI crawlers vs human users so you can A/B test AEO changes without regressing human UX.
Overview
Edge AI Delivery solves a common tension: the content that ranks well in AI engines is not always the content that converts well with human visitors. Long structured FAQ blocks, JSON-LD schema, explicit entity definitions — these help AI engines but can clutter a human-facing page.
Edge Delivery lets you define variants of a page and serve different variants depending on who is asking. AI crawlers get the AEO-optimized variant, humans get the human-optimized variant, and you can A/B test which variant performs better without deploying conflicting content on the same URL.
Plan gate: Edge AI Delivery is an Enterprise-only feature.
How it works
- You define one or more variants of a page — typically a "control" and an "AEO-enhanced" version
- Each variant has a targeting rule — user-agent match, IP range, header, or crawler identity
- The platform generates a small edge config you deploy to your CDN (Cloudflare Workers, Vercel Edge Middleware, Fastly, or similar)
- Incoming requests are routed to the correct variant based on the rules
- The platform tracks which variant AI engines were served and correlates it with visibility changes
The edge config is a JSON file with routing rules — no code deployment required on the AEO Optima side.
Configuration
Open Edge Delivery from the sidebar. You'll see:
- A list of configured URLs
- For each URL: the variants, targeting rules, current status, and A/B test results if applicable
- A Create variant button for each URL
To set up your first variant:
- Click Create variant
- Enter the target URL (must be a URL you own and track)
- Paste the variant HTML (or the rewritten copy — the platform wraps it in the page template)
- Pick a targeting rule: by default, all known AI crawler user-agents are matched
- Save
- Follow the deployment instructions specific to your CDN
Measuring variant performance
After a variant is live, the platform tracks:
- Which variant each AI engine was served (based on crawler logs)
- Visibility score for prompts citing the URL, segmented by variant
- Whether visibility improved, regressed, or was unchanged after the variant went live
You'll see these metrics on the variant detail page and they feed into Trends and Reports.
Best practices
- Test one variable at a time. Running multiple variants with different changes makes attribution impossible.
- Run tests for at least 2-4 weeks. LLM indexing cycles are slow — short tests rarely show statistically meaningful results.
- Keep a control variant. Always have one variant identical to your current page as a baseline.
- Don't serve contradictory facts. Different variants should optimize presentation, not rewrite claims. Conflicting facts across variants hurt your entity clarity score.
- Monitor crawler logs. The Crawler Intelligence page shows which crawlers are hitting each URL and helps you verify your targeting rules are working.
Related docs
- Crawler Intelligence — see which AI crawlers are hitting your site
- Content Optimization — generate the AEO-enhanced variant copy
- Schema Builder — generate the structured data to embed