A comparison. Not a verdict.

SmartBidCheck reads contractor bids with the same checklist every time. Company names do not change the checks. The comparison rules are the same for every contractor. The footer notes who built the tool.

What the tool does

Claude reads the uploaded pages and extracts stated scope, prices, allowances, exclusions, and terms into the checklist below. It also unpacks explicit tasks inside scope paragraphs and product packages. Software aligns those tasks using the same component and action vocabulary for every bid, then generates questions from document facts. There is no second AI judgment that chooses a winner.

Each finding includes source wording when available. The tool does not verify the truth of that wording, the contractor’s qualifications, product quality, or whether a price is fair. Check the original document, especially where the text is unclear.

Every bid uses the same extraction prompt and deterministic comparison rules. Company identity is only a label. No weighted criteria, scores, rankings, or special exemptions.

Same job does not always mean the same starting point

Supply, removal, new installation, and reinstallation are separate commitments. A contractor may be reusing a tub that another bid originally supplied. A brief package may explicitly include several tasks without pricing each one. Those written tasks are extracted. Work-detail rows do not allocate prices at all; original prices stay in quoted line items and the trade checklist. A package price is never split into estimated costs or deducted as a presumed product-only cost.

Matching task labels do not prove equal quantities, dimensions, materials, products, or labor. The details and original wording remain visible. An installation statement alone does not establish who supplies the product, and industry expectations are not treated as written commitments. “Not documented” means confirm, not assume missing work.

Optional project context is entered after results. It is clearly labeled as information you supplied, separate from bid evidence. It does not change document statuses or adjust quoted totals, and is included in exports and saved reports when you choose those actions. More pages or more wording do not earn a bid a better score; there are no scores.

What the labels mean

Included: the document makes an explicit commitment. If no amount is stated, it is not treated as zero.
Allowance: a provisional budget, with details and amounts as stated. Product selections or overages may change it.
Excluded: the document explicitly leaves something out. It is shown separately from missing wording.
Not mentioned: no wording was found in the uploaded pages. It may not apply to the project; it is not assumed missing from the actual work.
Needs clarity: wording is vague, conditional, conflicting, TBD, or difficult to read.
Not in this scope: the document explicitly says a category does not apply.

“Check source” indicates low extraction confidence, not a judgment about the bid. Category amounts may overlap, include other work, or represent allowances: they are never summed into an invented total. Stated totals are preserved. Arithmetic is checked only when all additive line items are explicitly captured.

Questions prioritize vague wording, allowances, explicit exclusions, categories documented in another bid, and contract terms. A trade absent from every bid is not automatically called a gap. Roofing, for example, may have nothing to do with a kitchen remodel.

Your bids are yours.

The privacy page covers retention, processors, and deletion. In short: source files are not saved by SmartBidCheck. Extracted report text is kept for 30 days from the last save, share, edit, or payment. Documents are sent to Anthropic only to be read. We do not use your bids to train any AI model. Anthropic’s commercial API terms do not use API inputs to train models. Anthropic’s standard policy deletes API inputs and outputs within 30 days, with exceptions for safety, legal obligations, or different contracts. This app does not claim zero retention by Anthropic. See Anthropic’s retention policy.

Sample bids are fictional. They are not real proposals and they are not current prices. Analytics does not receive bid contents or share links.

Limits worth knowing

AI extraction can miss text, misread numbers, or misunderstand context. Bad scans, handwriting, complex layouts, missing pages, and ambiguous totals are especially difficult. This version accepts 2–4 bids, up to 10 MB and 60 PDF pages each. Larger PDFs are split in your browser; large scanned pages may be converted to images before reading. Processing time varies with document length and service load.

Always verify license, insurance, references, final scope, and payment terms directly. This is document assistance, not construction, legal, or hiring advice.