How salary estimates work

When a listing includes a salary range, winq shows it exactly as posted — bold, no prefix. When a listing doesn't include pay, winq shows an AI estimate prefixed with a tilde (~) and a dotted underline. These are two different things. The visual difference is intentional.

Estimates are written by the AI model that scores your jobs (Gemini 3.8 Flash unless you chose another; on the free first search, before you add a key, DeepSeek V4.1 Flash) when the posting lists no salary. It reads the posting, mainly the job title, the company and the location, and draws on its training data — which includes compensation surveys, job listings, and public salary data indexed before its knowledge cutoff.

For well-known companies with standard roles in major markets, estimates tend to land in the right ballpark. The AI has seen a lot of software job data.

Estimates don't reflect equity, bonuses, or total compensation — only base salary. They don't account for the specific team budget, recent funding, or internal pay bands that the hiring manager knows and you don't. Estimates for early-stage startups, niche roles, or non-US markets are less reliable. The AI has no visibility into what this specific company is actually willing to pay right now.

winq asks the AI to self-assess its confidence: low, medium, or high. High confidence typically means the role is common and the company is well-documented. Low confidence usually means the role is unusual, the company is small or obscure, or the location has limited comparable data.

Confidence is AI self-reported and imperfect. A "high confidence" estimate can still be wrong. Treat it as a rough signal, not a guarantee.

Salary is one of the most useful filters in a job search, and most postings hide it. A rough estimate — clearly labeled as such — is more useful than nothing when you're deciding whether a role is worth pursuing. You already know the posting is silent on pay; having a directional signal lets you prioritize triage.

We're not trying to anchor you to a number. We're trying to help you spend your time on roles that are likely in your range. Always verify in the actual conversation.