How AI reads Portillo's

portillos.com Jun 30, 2026 6 min read Basic Web Presence

Short

39/ 100

AEO Level 2Basic Web Presence

The short answer

Portillo's earns a 39/100 AEO score at Level 2 readiness. Structured data shines at 80/100, helping answer engines surface hours and menu details, but agent interfaces and identity auth both sit at zero, leaving AI assistants unable to interact with the site beyond passive reading.

What AI sees

An AI agent visiting Portillo's today walks away with solid restaurant facts but no path to programmatic engagement.

When an AI crawler lands on Portillo's homepage it encounters well-formed structured data covering restaurant locations, menu categories, and operating hours—strong signals for answer engines fielding questions about Chicago-style Italian beef or chocolate cake. Content structure scores a moderate 40, reflecting readable headings and descriptive copy. Everything beyond that passive layer is absent: no Markdown response path, no API catalog, no MCP Server Card, and no explicit AI crawler rules in robots.txt. An agent trying to do more than read static facts is essentially cupping its hands against the glass rather than stepping inside.

Portillo's homepage as an AI crawler encounters it—structured restaurant data present, no agent interface layer to go deeper

Portillo's robots.txt showing no AI crawler directives—GPTBot, ClaudeBot, and PerplexityBot are operating entirely on default assumptions

Where it loses points

Agent interfaces and identity auth each score zero—the sharpest drag on a brand whose customers ask AI assistants about menus and drive-through locations every day.

How to fix it

Three targeted changes move Portillo's from a site AI agents can only observe to one they can genuinely use.

Declare AI crawler rules in robots.txt

Goal

Give major AI crawlers explicit permission rules so they know exactly which paths they may index and cite.

Issue

The scan found no user-agent entries for GPTBot, ClaudeBot, PerplexityBot, or any other AI crawler in robots.txt.

Fix

Add per-crawler User-agent blocks with Allow directives for menu pages, location pages, and any content worth surfacing in AI answers; apply Disallow to checkout and account paths. This single file edit governs how every major answer engine ingests the site.

Serve Markdown for agent requests

Goal

Return a Markdown version of page content when an AI agent signals it prefers plain text over HTML.

Issue

The site returns standard HTML even when a request carries Accept: text/markdown, giving agents a noisy document to parse.

Fix

Add server-side content negotiation: when the incoming Accept header includes text/markdown, respond with a clean Markdown document and set Content-Type: text/markdown. Browsers keep receiving normal HTML while agents get a structured, parseable version ideal for citation in AI-generated answers.

Publish an MCP Server Card

Goal

Let AI agents discover Portillo's capabilities through a standard machine-readable card at a well-known path.

Issue

No MCP Server Card was found at /.well-known/mcp/server-card.json.

Fix

Create and serve /.well-known/mcp/server-card.json containing serverInfo, a transport endpoint, and declared capabilities such as menu lookup or nearest-location search. This unlocks integration with MCP-compatible AI clients that can answer customer questions—like finding the nearest Portillo's or current seasonal items—dynamically rather than from stale training data.

Common questions

Why does Portillo's score 39/100 even though structured data rates 80?

Structured data at 80/100 helps answer engines surface static facts—restaurant hours, addresses, menu categories—but AEO scoring also weights agent interfaces and identity auth, both at zero. Without Markdown negotiation, an API catalog, or an MCP Server Card, AI agents cannot interact with Portillo's beyond reading, which pulls the composite score down to 39.

What does an agent_discovery score of 60 mean for Portillo's?

Agent discovery measures how easily AI crawlers can find and interpret a site's structure. Portillo's reaches 60 through its structured data foundation, but missing robots.txt AI crawler rules and absent RFC 8288 Link response headers leave real gaps. Closing those two issues alone would push this category noticeably higher and improve how confidently AI assistants cite the brand.

How long would it take Portillo's to meaningfully raise its AEO score?

The three highest-impact fixes—robots.txt AI crawler rules, Markdown content negotiation, and an MCP Server Card—are back-end changes a development team could ship in a single sprint. The robots.txt edit takes minutes; Markdown negotiation and the Server Card require a few hours of configuration. Together they could realistically add 20 to 30 points to the overall score.