Search does not read the page you are looking for. It reads a copy of that page's text, written when the work was published and rewritten whenever the work changes. If a word is not in that copy, nothing comes back for it, however plainly the word sits on the page itself.
What each result type puts into that copy
- Projects. The title and description of the published version, the discipline tags, the first name, last name and username of every team member, the location down to city, province and country, and text read out of the project's own images.
- Journals. The title, the tagline, the opening 1,000 characters of the article, and its disciplines and tags.
- Competitions. The title, tagline, disciplines, the organizer's name, the opening 1,000 characters of the About text, every award's name, amount and description, the location, the entry type, the jury members' names and the partners' names.
- Publications. The title, description and disciplines. When the publication is a PDF, the text of the pages is included as well.
- Topics. The title, description and topic tags.
- People. First and last name, username, and the bio and summary on the profile.
- Organizations. The title, tagline and description.
- Market. The product name and the group it sits in.
- Images. Any title on the file, the words readable inside the picture, a written description of it, and its tags.
Images are searched by what is in them
Every image is read once by an automatic analysis. It pulls out the text visible in the picture, writes a short description of what it shows, produces up to 30 descriptive tags, picks 5 dominant colours and files the image under one content type: floor plan, section, elevation, 3D render, site plan, diagram, presentation board, collage, photo or other. Those tags and that description are what your words are matched against, which is why an untitled drawing is still findable.
For a project, that same image text is folded into the project's own entry. So a project can match on a word that appears nowhere except inside one of its drawings.
What search does not cover
Help articles are searchable too, and appear under Help. Comments are not indexed, and neither are direct messages or the resource files attached to a project. Work that is unpublished, private or deleted has no entry at all. Only the published version of a project is copied, so an edit you have not published yet cannot be found. Hidden profiles are left out.
Your words are then matched two ways at once, literally and by meaning. How search ranks results explains which of the two decides the order.