How it works

The resume and the LinkedIn profile go to two independent model calls that never see each other. This is the whole reason the cross-check means anything: a single pass over both documents would quietly reconcile the differences, and the discrepancies it then reported would be the ones it chose to keep.

Each pass is transcription only. It copies employers, titles, and dates as written, quotes the line it read each role from, and records anything it could not make out rather than guessing.

Tenure lengths, the gaps between roles, overlapping employment, internal promotions, and every date comparison between the two documents are calculated in code from the transcribed dates. Language models are unreliable at arithmetic and very willing to produce a confident wrong number, so they are not asked to.

The figure that carries the most weight is the current tenure divided by the candidate’s own median completed stint. A candidate’s history is the only fair baseline: four years is a short run for one person and a long one for another, and comparing anyone against a general notion of “normal” tenure says more about the notion than the candidate.

The verdict is about job-change propensity — how ready this person appears to be to move roles at all. It is not a claim about fit for your role, and not a prediction that they would accept your offer.

Those would need a job description, a compensation band, a location, and a company profile, none of which this tool asks for. Any product that claims otherwise from a resume alone is guessing.

People update one profile and forget the other constantly. A title that differs between a resume and a LinkedIn page is far more often a stale page than a false claim. The report ranks differences by how much they are worth asking about and gives you both quotes side by side so you can ask.

The model is instructed to reason only from work history, tenure, titles, and career trajectory — and never from age, gender, ethnicity, nationality, religion, caste, marital or family status, disability, or appearance, including anything that might be inferred from a name, a graduation year, or a place of study. University and employer prestige are not treated as proxies for anything.

Every signal in the report has to cite the fact or statistic it rests on. Signals that could not be grounded are dropped rather than guessed at, and a thin document produces a low-confidence verdict instead of a padded one.

Resume Trust never connects to LinkedIn, logs into it, or fetches anything from it. Automated scraping breaks LinkedIn’s User Agreement and gets accounts and addresses banned. You supply the profile — LinkedIn’s own More → Save to PDF export, or pasted text — and the tool reads what you give it.