Why we label at all
AI-generated content and AI-assisted recommendations are quietly everywhere now, and often indistinguishable from the human or purely-algorithmic version — on purpose, in a lot of products, because ambiguity performs better on engagement metrics than honesty does. We think that trade is backwards. "AI always labeled" is one of the five short promises in our manifesto, and it applies in two directions: to content we publish ourselves (images, illustrations, copy), and to recommendations you see inside AlignAlgo's tools, once those tools are live and making judgment calls on your behalf.
Our provenance taxonomy
Rather than a simple "AI or not" toggle, we use a five-way taxonomy built to be honest about uncertainty rather than to overclaim detection certainty:
Made by a person, without AI generating the underlying content.
The content itself — an image, a block of copy — was produced by an AI system.
A person made it, with AI helping somewhere in the process — drafting, editing, compositing.
Backed by cryptographic or metadata proof (e.g. C2PA) of how something was made, not just a claim.
We genuinely don't know. This is an honest, common answer — not a failure state.
The rule underneath all five: declared is not the same as verified, and verified is not the same as inferred — we say which one applies. We never claim human-made certainty from a detector's guess, and "Unknown" is a real, acceptable answer rather than something to paper over.
Where this applies today
Being phase-appropriate here matters more than sounding finished. The full five-way taxonomy is a shared engine still maturing as a Creytix capability, not something stamped as a badge on every asset across this site yet. What exists today is a simpler, plainer-language version of the same discipline: image credit lines under article hero images. For example, the hero image on the no-ads manifesto article is credited "Composed from Align brand assets (not AI-generated)" — a deterministic composition, not an AI generation, and the caption says so directly rather than leaving it ambiguous.
The homepage itself is built almost entirely from CSS, SVG, and hand-composed app mockups rather than photography, so there isn't currently a large surface of photographic or AI-generated imagery on alignalgo.com that would need a five-way badge. As product surfaces launch — Align Shop's trust scoring, Align Food's taste matches, Communities' recommendations, and the rest of the family described on the homepage — we expect this same discipline to extend from article credit lines into per-result labels inside the live tools.
What we don't yet claim
- We don't run a detector that certifies "definitely AI" or "definitely human" for arbitrary content across the web. When we label something Unknown, we mean it.
- Verified provenance (cryptographic or metadata proof, like C2PA) is only as good as what the original publisher attached — we'll say when we have that proof and when we're inferring instead.
- This page describes AlignAlgo's own content and, going forward, its own recommendation surfaces — it isn't a claim about labeling practices on other sites or platforms.
How this ties to recommendations
This disclosure sits next to two other manifesto promises: "See every parameter" and "Ads only on your terms," both explained in the no-ads manifesto article. Once AlignAlgo's recommendation surfaces are live, if an AI system made a judgment call in a ranking or a match, that involvement is disclosed too — not folded quietly into "the algorithm decided."
Questions
Questions about how something on this site was made, or about this disclosure generally, can go to [email protected].