Why people pay more for human-made
The research is more mixed than either side claims. Labels raise perceived authenticity and can also raise perceived novelty, what matters is which one your category rewards.
Handmade goods have carried a price premium for as long as there has been machine production. The interesting question is not whether that premium exists but whether it transfers to digital work, where the object is identical either way.
The honest answer from the research is: sometimes, and the direction depends on the category.
What the research actually shows
Peer-reviewed work on AI disclosure has grown quickly, and it does not point one way.
A 2025 study in Psychology & Marketing found that framing changes willingness to pay, describing a visual as “created” by AI rather than “generated” by AI raised what people would pay, because the framing raised perceived value. The disclosure was constant; only the verb changed.
Later work in 2026 on AI disclosure labels found the effect splits into two opposing forces. Labels can raise perceived novelty, which improves attitudes and purchase intent. Labels can also lower perceived authenticity, which does the opposite. Which force wins depends on the product type.
Research from Wharton on generative AI and artists examines a related question, consumer preferences around artistic style and fair compensation, and again finds attitudes contingent rather than uniform.
A widely circulated claim puts the effect at “32% less creative, 41% lower willingness to pay.” Those figures appear in secondary summaries with no traceable primary study, so they are not used here.
Where the premium is real
The pattern across this literature is that the premium is strongest where the maker is part of what is being bought:
Commissioned and personal work. A portrait, a wedding photograph, a piece made for one person. The value is partly relational; a generated version is not a cheaper substitute, it is a different product.
Work sold with a story. Craft markets have always priced the maker’s hands into the object. Digital work sold this way behaves the same.
Signed and limited work. Scarcity is a human-made property. Generation is unlimited by construction.
Work where taste is the product. Editorial illustration, art direction, brand design. The client is buying judgement, and judgement is the thing they cannot verify from the artefact, so provenance stands in for it.
Where it is weak or absent
Commodity output. Stock imagery, boilerplate copy, background music. If the buyer’s requirement is “something that fills this slot,” provenance is not part of the specification.
Anywhere the buyer never learns. A premium requires disclosure. No label, no effect, which is the entire commercial argument for labelling.
Categories where novelty beats authenticity. The 2026 disclosure research is explicit that this happens; advertising is one documented case.
The mechanism underneath
Three things seem to be doing the work: effort (people value what took human time), intent (a choice made by someone rather than sampled from a distribution), and relationship (someone specific is accountable for it).
Notice that none of these are properties of the artefact. They are properties of its history. Which is exactly why provenance and labelling matter commercially, not just ethically, the premium cannot attach to something the buyer cannot see.
The practical reading
If you sell into categories where the maker is part of the product, disclosure is worth something and worth making legible. If you sell commodity output, it is not, and no amount of labelling changes that.
Anyone claiming a universal human-made premium is overselling the evidence. Anyone claiming there is none is ignoring it.