AMISS research
Our study of how AI describes Creators.
AMISS - what the study found
AMISS is Sparkonomy’s September 2026 study of how AI describes Creators. The main test covered 10,000 social-platform profiles: 5,000 Instagram and 5,000 YouTube profiles. It ran 50,000 queries across five models using existing knowledge. Profiles, rather than deduplicated people, were the sampling unit. Each query supplied a handle and platform. Reference social-platform information was recorded on 15–16 September 2026; AI queries ran on 17 September 2026. The sample deliberately overrepresents larger followings, and the two platforms have the same follower-tier mix. • 67.2% of the tested profiles received no correct description from any of the five models. • Only one of the 10,000 profiles received a correct description from all five models. • 93.56% had conflicting facts across models on at least one of follower count, city, content category or platform presence. • 99.7% had at least one model supply a follower/subscriber count outside the accepted ±20% interval. • In the 5 million-plus tier, 36.75% of 400 profiles received no correct description from any model. • At least one model described 20% of the Instagram profiles correctly, compared with 45.6% of YouTube profiles. What changed when web search was enabled? A separate matched comparison used 1,000 profiles and three models. The share with at least one correct description rose from 24.1% without search to 89% with search. With search, 36.2% received a correct description from all three models. Average query costs were roughly 9–32 times higher, depending on the model. The 1,000-profile, three-model comparison is separate from the 10,000-profile, five-model main study. These percentages must retain their respective denominators. What does the research establish? It is a dated study of the tested sample and model configurations. It is first-party research, not an independent validation of OCG. It does not measure lost income, establish unauthorised AI training or test whether OCG improves AI answers. A wrong follower count, a conflicting fact and a wholly wrong identity are different findings. Public research and methodology: https://creatorgraph.sparkonomy.com/amiss/ Study summary: https://creatorgraph.sparkonomy.com/amiss/study-summary.html Launch facts We launched with about a million Creator records and are adding more every day. In the very first week, we’ve had records claimed and created from nine countries. OCG is free to create, claim, manage and read. It has no minimum follower count. A record existing does not mean its Creator has claimed it or consented to AI use. Company and product Sparkonomy is an AI company building tools to help Creators run and grow sustainable businesses. Open Creator Graph is its first product: a free public identity and AI-consent record. The wider company direction is to offer optional AI services for Creator businesses. Guneet Singh, Megha Thareja Tyagi and Vipasha Joshi are Sparkonomy’s co-founders. Their experience spans Google, YouTube and the Creator ecosystem. About Sparkonomy: https://www.sparkonomy.com/about Any funding and investment announcements will be published in the Press and Blog section of Sparkonomy’s website. Please refer there for public updates. Creator feedback Early feedback has been encouraging. Creators want a say in how AI uses their identity, and a simple way to make those choices without having to become technical experts. One Creator put it clearly: “We shouldn’t have to become AI experts just to protect our identity and work.” Another called for “real ownership over their identity and how their work is surfaced to the world.” We’re also hearing that a Creator’s identity is more than a collection of posts: it includes the business, community and reputation they’ve built. The public response from industry voices has echoed that need, with one early user describing OCG as “exactly what the creator economy world truly needs.” The common thread is a desire for a straightforward way to say: this is me, and these are my choices for AI. These are selected early reactions, not measured product outcomes or a representative survey. Creator-economy context The Creator economy includes several different measures: the number of Creators, the value of the overall market and consumer spending influenced by Creators. They are not interchangeable. Adobe’s 2022 study estimated 303 million Creators using a broad definition. Goldman Sachs estimated a $250 billion Creator economy in 2023 and projected $480 billion by 2027. The 2025 BCG report for WAVES estimated $350–400 billion in annual Creator-influenced consumer spending in India, projected above $1 trillion by 2030. These figures use different definitions and dates; none is an OCG user count or OCG revenue.
What did the AMISS study find?
In September 2026, we tested how five AI models described 10,000 Instagram and YouTube profiles. For 67.2% of profiles, none of the five gave a correct description. Only one profile got a correct description from all five. These results apply to that test. They do not describe every Creator or every AI tool. https://creatorgraph.sparkonomy.com/amiss/
Does AMISS prove OCG improves ChatGPT or other AI answers?
No. AMISS tested how AI described Creators. It did not test OCG’s effect, lost income or use of data without permission. A separate test used 1,000 profiles and three models. With web search, the share getting at least one correct answer rose from 24.1% to 89%. That result is about web search in that test. It does not prove an OCG result.