AI Research Specialist Needed -- 2
Publicada el 2026-07-22
Descripción de la oferta
Remote AI Research Tasker We are looking for a detail-oriented, self-motivated individual to join an AI research project. In this role, you will write challenging, multi-source research prompts designed to test the limits of AI models, helping them get better at searching the web over time. What You'll Be Doing You'll research specific verifiable facts across multiple public websites, craft precise natural-language questions paired with a single correct answer, and document a detailed step-by-step browsing path showing exactly how you found it. Every task goes through automated and human review before payout, so quality and accuracy are everything. Key Responsibilities Write research-heavy prompts (70–150 words) that require at least 3 distinct public web sources to answer Pair each prompt with a single atomic answer — a word, name, number, date, or short phrase Build a thorough step-by-step browsing path documenting the exact URLs and page locations used to find each fact Test every prompt in ChatGPT 5.5 Pro and honestly record model responses Run through a full pre-submission checklist before every task Revise and resubmit tasks based on reviewer feedback when needed Skills & Requirements Research & Sourcing Strong ability to find specific verifiable facts across multiple public websites Skilled at identifying authoritative, paywall-free sources Able to cross-check facts across sources to confirm one single defensible answer exists Knows how to test links in incognito mode to confirm public accessibility Critical Thinking & Logic Can craft questions with exactly one unambiguous answer — no yes/no, true/false, or binary responses Able to identify when a prompt is too vague or leads to more than one valid answer Understands how to build constraints that genuinely narrow results down to a single fact Can think like a reviewer and spot weaknesses in your own work before submitting Writing Writes clear, natural-language questions free of process instructions, vague superlatives, or floating time words like "currently" or "latest" Produces original prompts only — no copying from trivia sites, question banks, or existing question sets Answers written in atomic format only — single value, no sentences or explanations Web Navigation & Documentation Can map out a precise step-by-step browsing path showing how each constraint in a prompt is resolved Able to document exact URLs, sections, table rows, or paragraphs where each fact is found Understands how search queries work and can describe exactly which result was clicked and why AI Testing Comfortable running prompts through ChatGPT 5.5 Pro and recording results accurately and honestly Able to judge whether a model failure is genuine or caused by a flawed prompt No cherry-picking runs, re-running for better results, or editing model responses Attention to Detail Able to work through an 18-point pre-submission checklist on every task Can catch and fix common rejection reasons including ambiguous wording, missing trajectory steps, broken source links, and incorrect intermediate answers Understands the full review pipeline from submission through to final approval Tech Must have or be willing to set up ChatGPT 5.5 Pro with correct configuration (memory off, correct model and effort settings) Comfortable working independently online Reliable internet and ability to manage multiple browser tabs at once Pay Task-based work: $10 per approved task Hourly work: $5 per hour Rejected tasks can be revised and resubmitted as many times as needed until approved No cap on how many tasks you can complete and get paid for
Skills
Fuente original: freelancer