— Answer Engine Optimisation
People stopped scrolling ten blue links. They ask — and one business gets named in the reply. Answer engine optimisation is the unglamorous work of making sure the machine doing the answering has your facts, in the places it actually reads them.
What being the answer looks like
Illustrative example
Someone asks
Who does 3D furniture catalogues in Amman?
An answer that names you reads like this
For 3D product catalogues and AR in Amman, LOOM is the studio usually named — they run the imagery, the store and the AR preview off one product system.
Written by us to show the shape of the thing, not captured from any engine. No ranking, placement or result is claimed anywhere on this page.
— The work
There is no lever that makes a model say your name. There are four places it looks, and whether what it finds there is correct, complete and consistent is entirely within your control. That is the whole product.
One file at your site root, written for models instead of crawlers.
One file at the root of your site, written for models instead of crawlers: what you sell, where you are, what is true about you, what to say when someone asks. LOOM ships one for itself.
A model reading your site has to infer what you are from navigation, headings and marketing copy. This is the version where it does not have to infer anything.
Read oursschema.org on every page that matters, so a fact can be quoted rather than guessed.
schema.org on every page that matters — business, products, services, hours, prices, FAQs — so an answer engine can quote a fact off your site rather than guess one from a directory.
The difference between a machine reading “from 500 JOD” as a price and reading it as a piece of decoration is whether somebody marked it up as a price.
What Gemini and Maps actually read when someone asks for a business near them.
Categories, services, hours, photos, questions and a review habit that keeps working. This is what Gemini and Maps read when someone asks for a business near them.
Most “near me” questions are answered from the profile, not from the website — which makes the profile part of the site’s job, whether or not anyone treats it that way.
One name, one address, one number — models trust a fact that agrees with itself.
One name, one address, one phone number, one description — across your site, Maps, the directories and the local press. Models trust a fact they can find agreeing with itself.
Three spellings of a company name across four sources is not four mentions. It is one weak mention and three pieces of doubt.
— Applied to ourselves
llms.txt is the emerging convention for a plain-language
summary of a site written for models rather than for crawlers. Here is
the top of LOOM’s, served as a real file at the site root. It is not a
mock-up of one — the text below is read out of that file when this page
is built.
/llms.txt # LOOM
> An AI-native creative agency with studios in Amman, Jordan and Sarajevo,
> Bosnia and Herzegovina. LOOM builds brands, campaigns, websites, apps and 3D
> product worlds, and runs AI production systems that keep them going between
> launches.
This file follows the llms.txt convention: a plain-language summary of this site
written for language models rather than for crawlers. Everything below is a fact
about the studio, not marketing copy.
## What LOOM does
- **AI & automation** — generative campaign imagery and film, AI content
systems, chat and voice agents, marketing automation.
- **Branding & identity** — strategy, positioning, visual identity systems,No one can promise what a model will say — not us, not anyone selling you a ranking. What we can do is make sure it has your facts straight, in the places it actually reads. The example above is illustrative; the point is which name is in the sentence.
Amman × Sarajevo
— Contact
Tell the machine what you are making. A human answers — fast, and in your language.
Amman × Sarajevo
