You have no website. Here is how an AI assistant finds you anyway.
Short answerAn assistant can only recommend what it can read, and an Instagram profile is close to unreadable to one. Tunnel builds the machine-readable record you never had — what you sell, where you are, when you are open, how to reach you — hosts it at a permanent URL, and serves it from the same endpoints a crawled website gets. You do not build or maintain anything. The free tier covers one business.
A visibility tool works by reading the website you already have, scoring it, and handing you a list of things to change on it.
You do not have one. So there is nothing to score — we build the record itself and host it at an address anything can fetch.
You don't need a website to be found
Plenty of good businesses sell entirely through Instagram, TikTok or a phone number. To an AI assistant those are close to invisible — there is nothing crawlable to read, so you never make it into the answer.
Tunnel builds and hosts the machine-readable record for you. It is the same listing a crawled website gets, minus the website.
You send us where you already are
An Instagram handle, a TikTok, a Facebook page, a phone number. That is enough to start.
We build the record you never had
What you sell, where you are, when you are open, how to reach you — written once, in the formats assistants read.
It lives at a real address
Your listing gets its own permanent URL and machine-readable endpoints. No site to build, nothing to maintain.
This is what a machine gets when it asks about you
Not a screenshot of a dashboard. The response from the endpoint your listing is served at, for a business whose only other address is an Instagram handle.
https://api.tunnelpowered.com/api/kb/entities/{slug}?format=jsonldexample{
"@context": "https://schema.org",
"@graph": [{
"@type": "LocalBusiness",
"@id": "https://tunnelpowered.com/business/brutaria-de-pe-colt#entity",
"name": "Brutăria de pe colț",
"url": "https://tunnelpowered.com/business/brutaria-de-pe-colt",
"areaServed": "Chișinău — Botanica, Centru",
"openingHours": "Tu-Sa 08:00-19:00",
"sameAs": ["https://www.instagram.com/brutaria.de.pe.colt"],
"makesOffer": [
{ "@type": "Offer", "name": "Tort la comandă",
"price": "450", "priceCurrency": "MDL" }
]
}]
}An excerpt — the live response is minified and carries more fields than this. Every name below is one the endpoint really sends, and nothing has been renamed to read better on a page.
- sameAs
- Your Instagram, stated as the same business, in the one property a crawler is built to follow. A link in a bio is a link; this is an identity claim.
- openingHours
- Emitted only when your hours match the schema.org syntax exactly. “Mon to Sat, 8 till 7” stays prose instead of becoming markup that looks valid and is not.
- makesOffer
- What you sell, with a price and a currency where you gave us one — as a list a parser can read, rather than a sentence it has to guess at.
- @id
- A permanent identifier on tunnelpowered.com rather than on a subdomain. This is the address you paste into a bio, and it does not change if you get a website later.
What the record actually contains
The same fields a crawled website produces, gathered from you once instead of scraped from pages you do not have.
Who you are
Trading name, what kind of business this is, the city and address if you have a physical one, the languages you serve.
What you sell
Your products or services in plain sentences, plus a pricing table if you give us prices. Written to be quoted, not to be scrolled.
When and where
Opening hours, delivery area, whether you take walk-ins. The questions assistants are asked most often about a local business.
How to reach you
Whichever public channels you choose to publish, plus an agent-messaging address so an assistant can open a conversation without being handed a phone number.
Which channels you have proved
A social profile you have proved you control is marked as proved. One you have only claimed is not. The record says which.
The machine-readable serialisations
JSON-LD, an FAQ marked up as FAQPage, a knowledge page, a quotable answer block, a sitemap listing them — and llms.txt, which we ship without claiming anyone reads it.
Where it lives, and who can read it
A record nobody can fetch is a document, not a listing. Yours gets an address and appears in the places an agent looks.
A permanent URL of its own
Your listing is a real page at a stable address. You can put it in a bio, on a card, in a message — the same way you would a website.
The knowledge-base index and search
Every listing is enumerated in a public index and reachable by name or topic through a search endpoint, without a key.
An MCP server
Any assistant or custom agent that speaks Model Context Protocol can search the index and read your record as a tool call.
Crawlers that do not run JavaScript
Everything above is served as plain text or JSON. There is no client-side rendering for a crawler to fail at.
What this does not do
It does not create demand. Being readable is a precondition for being recommended, not a substitute for being worth recommending.
It does not put you into an answer. No assistant is obliged to retrieve your record, and we cannot make one. What we can do is remove every technical reason not to, and then measure what each assistant actually says about you.
It does not replace your social accounts. Keep posting. The record is the machine-readable copy of facts your Instagram states in pictures.
It does not answer for you. When an agent messages your business, a person still has to reply.
Questions people ask about this
Can I really be listed with no website at all?
Yes. An Instagram handle, a TikTok, a Facebook page or a phone number is enough to start. We build the record from what you tell us and host it ourselves, so there is no site to build and nothing on your side to maintain. One step is not instant, and you should know about it before you sign up: because there is no domain we can check by machine, a person at tunnel reads the profile before it is published anywhere — the web page, the directory, the search API and every assistant that reads us. Editing it afterwards sends it back for review, because the review was of the words that were there at the time.
What if I get a website later?
Add it as a website in the same account and it runs the normal pipeline: you put one integration line on it, we crawl it, and it gets six endpoints of its own. Your listing keeps its address — the profile URL is built from your business, not from your domain — so anything already pointing at it keeps working. Be clear that these stay two records rather than one turning into the other: the profile keeps the facts you entered, and we do not overwrite them with whatever the crawler found.
Does the free tier cover this?
Yes. The free tier covers one business, runs the full pipeline, and has no card and no expiry. That includes the hosted record, the six endpoints, the automated channel checks, a re-crawl every 30 days, and one visibility check — run when you join, and not repeated. Repeat checks are where the paid plans begin: every 7 days on Pro, every 3 on Business. A check means putting five questions to five assistants with web search on, which costs us real money every time, so the free tier gets one real check rather than a watered-down one on a loop.
How do you know the information is true if there is no website to check it against?
For the most part we do not, and the record does not pretend otherwise. What we can prove automatically is that you control the channels the record points at — a token in your public social bio, a code sent to your listed inbox. Two of those proofs are required before the record is marked as automatically verified, and they are re-checked on a schedule. Identity and whether the service is real are only established when a person checks them, which happens on paid plans.
Will an assistant recommend me over a competitor with a big website?
We cannot promise that and would not believe anyone who did. Retrieval and ranking happen inside systems we do not control. What being readable changes is that you become a candidate at all, in cases where previously there was nothing about you to retrieve.
How is this different from a tool that tracks my AI visibility?
A tracker reads what is already published about a business and reports on it. That is a useful thing and we do it too, on your plan’s schedule. But it needs something to read. If your business exists only as an Instagram profile, a tracker measures an absence and hands you a list of things to go and publish somewhere else. This builds the record itself, hosts it at an address of its own, serves it from endpoints that need no JavaScript, and then measures that.
Find out what AI says about your business
We ask four assistants what they say about your business, and send you the answers they gave — including the parts that are wrong. Free, and yours to keep.
Get your free AI visibility scan
By submitting, you agree to our Terms of Service and Privacy Policy