Manual screening and screening software apply the same criteria; the differences are coverage, consistency, and record-keeping. Software evaluates every ticker against every rule on every scan and logs why each candidate passed or failed. What software cannot do is decide what to trade: the judgment, and the decision, stay with the trader.
A disciplined manual pass over one candidate covers checking the technical setup, reading IV rank and comparing implied against realized volatility, confirming the earnings date, walking the chain for open interest and quote width at the candidate strikes, scoring expirations, computing the credit-to-width ratio, and sizing from worst-case loss. Done honestly, that is many minutes per ticker, which in practice caps a manual universe at a handful of names, and the checks drift on busy days. The drift is the expensive part: the criterion skipped once silently stops being a criterion.
It cannot know a thesis, a tax situation, a risk tolerance, or the difference between a statistically cheap stock and a business in decline. A screen describes what its rules found; it does not know whether the trade belongs in a particular portfolio. Screening software also inherits every limit of its own rules; a criterion nobody encoded is a criterion nobody enforces, and models mislead in regimes they were not built for.
Rules do the repetitive evaluation and the honest bookkeeping; the trader owns the judgment and every decision. That boundary is not a limitation of current software so much as the design: a screen whose output is a described set of facts, with reasons attached, leaves the reader free to disagree with it, and the record to check who was right.