Is your business visible to AI assistants?
A business is visible to AI assistants when its information is machine-readable and its services can be booked programmatically. In practice that means structured data on every page, an llms.txt summary, an agents manifest, a complete business profile, and a bookable endpoint an assistant can reach without a phone call.
A growing share of buying decisions now start with a question to an assistant rather than a search box. Someone asks ChatGPT for a dentist that does emergency appointments, or asks Perplexity to compare local HVAC contractors, and acts on the answer.
That channel behaves nothing like Google. It does not show ten blue links; it returns a short recommendation. If you are not in it, you are not runner-up - you are absent.
Why this is different from SEO
| Search engines | AI assistants | |
|---|---|---|
| Output | A list of links | A short synthesised answer |
| Positions available | Ten or more per page | Often two or three named options |
| What wins | Rankings and backlinks | Being parseable, citable and unambiguous |
| Fed by | Crawled pages | Crawled pages, structured data, and third-party roundups |
| Failure mode | Rank lower | Not mentioned at all |
The critical difference is the last row. On Google, weak SEO means less traffic. In an assistant answer, being unparseable means you simply do not exist in the response.
How to check your own visibility in five minutes
Do this now, before reading further:
- Open ChatGPT, Claude, and Perplexity.
- Ask each: 'What are the best [your service] providers in [your city]?'
- Ask: 'Tell me about [your business name].'
- Note whether you appear at all, and whether the details returned are accurate.
Most owners doing this for the first time find one of three outcomes: they are absent entirely, they are described inaccurately from outdated sources, or they are mentioned but with no way for the assistant to book them.
The five layers of agent readiness
1. Structured data
Schema.org JSON-LD on every page tells a machine what your business is, what you offer, where you operate, and when you are open. Without it, assistants parse unstructured HTML and often get it wrong.
2. An llms.txt file
A plain-text summary at yourdomain.com/llms.txt describing your business, services, coverage area and booking link. It is the equivalent of a business profile written for machines rather than people.
3. An agents manifest
A structured agents.json (and a copy at /.well-known/agents.json) describing your capabilities and how an agent should contact or book you. This is what tool-calling assistants read when they want to act rather than just describe.
4. A bookable endpoint
This is the one most businesses fail. If the only way to book you is to call a phone number, an assistant cannot complete the task - so it recommends a competitor who can be booked with a link. A scheduling deep link is the minimum viable answer.
5. Third-party presence
Assistants lean heavily on roundups, directories and review sites when recommending vendors. If the 'best X in Y' articles for your category do not mention you, you are missing from a primary source these systems quote.
A practical readiness checklist
| Check | Why it matters |
|---|---|
| JSON-LD on every page | Machines read facts instead of guessing from HTML |
| /llms.txt exists and is current | Gives assistants an authoritative summary |
| agents.json + /.well-known copy | Enables tool-calling agents to act |
| AI crawlers allowed in robots.txt | Blocking GPTBot or ClaudeBot removes you entirely |
| Bookable link, not just a phone number | Lets an assistant complete the booking |
| Business profile complete and verified | Feeds local recommendation data |
| Present in category roundups | Assistants cite these when recommending |
| Content answering real questions | Provides quotable material |
The robots.txt trap
Many sites block AI crawlers by default, sometimes through a security plugin or CDN setting nobody reviewed. If GPTBot, ClaudeBot, PerplexityBot or Google-Extended are disallowed, no amount of structured data will help - you have opted out of the channel. Check this first; it is the single most common blocker we find.
Why acting early matters here
This channel is early enough that the competitive bar is low. Most businesses in most categories have none of the five layers above. That will not last - and assistants develop recommendation patterns that are sticky. Establishing presence while the field is empty is materially cheaper than trying to displace an incumbent later.
See your own numbers in about a minute.
The free calculator models your leak across five sources — no signup.
Open the calculatorFrequently asked questions
What is AEO?
Agent Engine Optimization is the practice of making a business discoverable, understandable and bookable by AI assistants. Where SEO targets ranking in search results, AEO targets being cited and actioned in AI-generated answers.
Is llms.txt an official standard?
It is an emerging convention rather than a ratified standard, and adoption is still growing. The cost of adding one is very low and it makes your business summary unambiguous for any system that does read it.
Do I need all five layers to benefit?
No. Structured data and unblocked AI crawlers deliver most of the initial benefit. A bookable link is the highest-value addition after that, because it changes you from a business an assistant can describe to one it can actually book.
How do I know if it is working?
Ask the assistants directly and repeat monthly. Query your category and your business name across ChatGPT, Claude and Perplexity, and record whether you appear and whether the details are accurate. There is no analytics dashboard for this yet, so manual checks are the practical measure.
Will this hurt my normal SEO?
No. Every component - structured data, clear content, a complete profile - is also a traditional SEO best practice. Agent readiness is additive rather than a trade-off.