Will a buyer really run my bid through an AI detector?

Some do, and the risk runs both ways: prose that reads as machine-written is a problem whether or not a machine wrote it. Knowing the score before you submit is the point, not proving authorship.

Does checking overlap mean you are accusing me of plagiarism?

No. The check reports where text matches material already published, with the source. Reusing your own past bid is normal and will still show as overlap. What matters is knowing before a buyer notices.

Tenders & Bidding•POST /v1/check-source-overlap

Check the response before a buyer does

Supported Formats: Draft tender responses, Capability statements, Method statements

Core Verification Mechanisms: Citation resolution and claim judgement · Published-source overlap · AI-writing signals

The problem

A response is a set of claims about certifications, experience, people and performance. Any of them can be asked about, and a claim nobody can evidence is a finding waiting to happen.

Text reused from a past bid or a template may overlap material already published on the web, and a buyer may run submissions through an AI-writing detector. Both read badly, and both are visible to the buyer before they are visible to you.

How Stipple helps

Resolve every citation in the response, then judge a sample of its claims — up to five in deep mode — against the sources they cite, so an unsupported assertion is found before submission.

Check the text for overlap with published material, with exact matches confirmed against their source URLs and paraphrase flags labelled as judgements rather than accusations.

Score the prose for how machine-written it reads, which is what a buyer running a detector will see whether or not a machine wrote it.

The call

bash

curl -X POST "https://www.stipple.sh/v1/check-source-overlap" \
  -H "Content-Type: application/json" \
  -d '{"text": "<your draft response>"}'
# -> { "matches": [ { "tier": "exact", "source_url": "https://...",
#      "matched_quote": "...", "source_title": "..." } ] }

The things a buyer would notice about your response are things you noticed first.

What the result looks like

An AI-writing score is about style, not authorship. It abstains where the text does not read as prose, which includes forms, tables and pricing.

Overlap is overlap, never an accusation. Exact matches are confirmed quotes with their source; paraphrase flags are labelled model judgements.

None of these is a compliance check or a prediction of how a response will score. They tell you what a reader might see.

The evidenceCitation hallucination benchmark: how often invented sources appear→

How the corpus was built, what was caught, and what the method does not prove.

Questions

Will a buyer really run my bid through an AI detector?: Some do, and the risk runs both ways: prose that reads as machine-written is a problem whether or not a machine wrote it. Knowing the score before you submit is the point, not proving authorship.
Does checking overlap mean you are accusing me of plagiarism?: No. The check reports where text matches material already published, with the source. Reusing your own past bid is normal and will still show as overlap. What matters is knowing before a buyer notices.

Sources

Check claims against sources