# Agent discoverability for SaaS sites

How ChatGPT agent, Perplexity Comet and Claude read your pricing, docs and sign-up, and what MCP, llms.txt and agent payments do today. We automate it.

Canonical: https://ignura.com/agents
Last updated: 2026-10-27
Author: Ignura

## In short

- Agents are AI tools that browse and act for a person: ChatGPT’s agent, Perplexity’s Comet, Claude in Chrome. They read pages the way a crawler does, then click and type.
- They already go looking for pricing. When we asked ChatGPT search which SOC 2 tool to buy, 7 of its 15 web searches were for pricing.
- Most of what helps agents is plain: prices in the page’s text, plans with clear limits, docs in HTML, a sign-up that doesn’t need a call.
- Some of it is new and unproven, like llms.txt and agent payments. We tell you which is which. Start with a free Growth Check.

## What an agent is today

OpenAI launched ChatGPT agent in July 2025: it uses its own virtual computer to navigate sites and asks permission before actions of consequence. Perplexity made its Comet browser free to everyone on 2 October 2025, with an assistant that acts on the page for you. Anthropic’s computer use, from October 2024, lets Claude look at a screen, move a cursor, click and type, and Claude in Chrome became generally available in August 2026.

It’s early. The agents are slow and make mistakes, and their makers say so. But they already read pricing pages and shortlist tools for people.

## Pages an agent can read

Most AI crawlers don’t run JavaScript (Vercel, late 2024). Agents that drive a real browser do, but they still work best from text in the page. Put prices, plan limits and who it’s for in plain HTML. Avoid prices in images, sliders that only show a number after a click, and “contact us” with no range.

schema.org’s Product and Offer give a price and its currency in a form any parser reads, and Google uses them for product results. Neither is required for AI answers, so we add them where they’re true, never as a trick.

**7 of 15** of ChatGPT’s web searches for our buyer questions were for pricing.

## Who gets in: robots, firewalls, signatures

robots.txt is a request, not a lock: Google says it can’t enforce crawler behaviour. OpenAI says robots.txt rules may not apply to ChatGPT-User, because a person asked for the page.

Firewalls matter more. Since July 2025 Cloudflare has asked every new domain at sign-up whether to allow AI crawlers. If yours says no, agents and AI search may never see you. OpenAI’s cloud browser signs its requests with Web Bot Auth, so your CDN can check it’s really ChatGPT and let it through.

## A sign-up an agent can finish

An agent stops where a person would get stuck: a demo wall, a phone field, a CAPTCHA. OpenAI’s help page says its cloud browser can’t sign in to websites or complete payments, so today the agent hands the last step back to the person. Make that step short: email sign-up, a trial without a sales call, one clear next action.

## MCP, APIs and llms.txt: early

The Model Context Protocol, introduced by Anthropic in November 2024, connects AI apps to tools and data. Its official registry of public servers opened in preview in September 2025 and is still in preview. If your product has an API, an MCP server is a real way to be used by agents. Whether it gets you chosen isn’t proven yet.

llms.txt is a 2024 proposal for a markdown guide at /llms.txt. Google's AI optimization guide says an llms.txt file "will neither harm nor help your site's visibility or rankings in Google Search". None of OpenAI, Anthropic or Perplexity says its crawlers read them. It's cheap, so we may add one. We won't sell it to you as GEO.

Agent payments are moving fast: OpenAI and Stripe launched the Agentic Commerce Protocol in September 2025, and OpenAI has since moved away from its standalone Instant Checkout. Google announced AP2 the same month. For B2B software, watch them; don’t build for them yet.

## Mistakes we see

- **Pricing that only says “Contact us”.** An agent can’t compare what isn’t there. Give a range or a starting price.
- **Prices that only exist in JavaScript.** Most AI crawlers read the HTML and stop.
- **A firewall that blocks every bot.** Check your CDN’s AI crawler setting before anything else.
- **Paying for “agent SEO” that is only llms.txt.** No major engine says it reads it. Cheap to add, nothing to pay for.

## What our AI does

- Fetches your pricing, plan and docs pages the way an AI crawler does, without JavaScript, and shows what it sees.
- Checks your robots.txt and CDN rules against each AI crawler and agent.
- Runs buying questions on ChatGPT and logs which of your pages it searches for.
- Walks your sign-up the way an agent would and records where it stops.

## What we do

- Rewrite pricing and plan pages so the facts are in plain text.
- Add structured data where it’s accurate.
- Shorten the sign-up path with your product team.
- Tell you which agent standards are worth your engineers’ time this quarter and which aren’t.

## The first four weeks

1. We read your site the way crawlers and agents do and list everything they can’t see.
2. Pricing, plans and who it’s for rewritten in plain HTML, with structured data where it’s true.
3. Firewall and robots rules fixed, the sign-up path shortened, docs checked.
4. We run the buying questions again and show which of your pages the engines now read.

## The deal

Named in 5 of your 20 buyer questions by week 4, or you don’t pay the second half. $8,000 for four weeks, half to start. Then $4,000 a month, with a month’s notice to stop.

## Questions

### Should we build an MCP server?

If agents could do real work with your API, yes: it’s a proven way to be used inside AI apps that support MCP. As a way to get recommended, it’s unproven. We’ll tell you which one you’re paying for.

### Do agents buy software on their own?

Not yet. OpenAI’s cloud browser can’t sign in or pay. Today agents research and shortlist, and a person signs up. That’s why being on the shortlist matters.

### Is llms.txt worth it?

It takes an hour. Google's AI optimization guide says an llms.txt file "will neither harm nor help your site's visibility or rankings in Google Search". No major AI company says its crawlers read it, so don't expect it to move anything.

### Will agent visits confuse our analytics?

They can look like normal visits. OpenAI signs its cloud browser’s requests, so your CDN can label them; we set that up where your tools allow, and cover the rest on our analytics page.

## Sources

- Ignura: [Six SOC 2 buyer questions on ChatGPT search, Perplexity and Google, 27 Oct 2026 (the data)](https://ignura.com/data/search-2026-10-27.json)
- OpenAI: [Introducing ChatGPT agent (17 Jul 2025)](https://openai.com/index/introducing-chatgpt-agent/)
- OpenAI: [ChatGPT Work’s Cloud browser allowlisting](https://help.openai.com/en/articles/11845367-chatgpt-works-cloud-browser-allowlisting)
- Perplexity: [Comet is now available to everyone worldwide (2 Oct 2025)](https://www.perplexity.ai/hub/blog/comet-is-now-available-to-everyone-worldwide)
- Anthropic: [Introducing computer use (22 Oct 2024)](https://www.anthropic.com/news/3-5-models-and-computer-use)
- Anthropic: [Claude in Chrome is now generally available (26 Aug 2026)](https://claude.com/resources/articles/claude-in-chrome-generally-available)
- Vercel: [The rise of the AI crawler (17 Dec 2024)](https://vercel.com/blog/the-rise-of-the-ai-crawler)
- Google Search Central: [Product structured data](https://developers.google.com/search/docs/appearance/structured-data/product)
- Google Search Central: [Introduction to robots.txt](https://developers.google.com/search/docs/crawling-indexing/robots/intro)
- OpenAI: [Overview of OpenAI crawlers](https://developers.openai.com/api/docs/bots)
- Cloudflare: [Cloudflare changes how AI crawlers scrape the internet at large (1 Jul 2025)](https://www.cloudflare.com/press/press-releases/2025/cloudflare-just-changed-how-ai-crawlers-scrape-the-internet-at-large/)
- Cloudflare: [Web Bot Auth](https://developers.cloudflare.com/bots/concepts/bot/verified-bots/web-bot-auth/)
- Anthropic: [Introducing the Model Context Protocol (25 Nov 2024)](https://www.anthropic.com/news/model-context-protocol)
- Model Context Protocol: [The MCP Registry, in preview (8 Sep 2025)](https://blog.modelcontextprotocol.io/posts/2025-09-08-mcp-registry-preview/)
- llmstxt.org: [The /llms.txt file (Jeremy Howard, 3 Sep 2024)](https://llmstxt.org/)
- Google Search Central: [AI optimization guide](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide)
- OpenAI: [Buy it in ChatGPT: Instant Checkout and the Agentic Commerce Protocol (29 Sep 2025)](https://openai.com/index/buy-it-in-chatgpt/)
- OpenAI: [ChatGPT for merchants](https://chatgpt.com/merchants/)
- Google Cloud: [Announcing the Agent Payments Protocol (AP2) (16 Sep 2025)](https://cloud.google.com/blog/products/ai-machine-learning/announcing-agents-to-payments-ap2-protocol)

Book a 15-minute call about your website: https://ignura.com/book
