The questions you should ask us.
Research is only worth what its method can defend. This page explains how n+1 collects, qualifies, and protects evidence, including where the method reaches its limits.
Who answers, and who doesn't.
Respondents are readers of premium publisher sites, reached by media targeting across geography, device, property type, and context. That produces engaged, in-market audiences rather than professional survey takers. It also means samples reflect the reach of the media plan, not a probability panel; we say so, and we scope claims accordingly.
Only genuine answers count.
Every message is classified for genuineness before it advances a survey; gibberish and abuse never count toward a completion, and abusive sessions are halted. A response qualifies only after three sufficient answers. Completed respondents are remembered for 90 days and not re-surveyed, and one response can bill at most once.
Always labeled. Incentives designed, not hidden.
Every unit is labeled "AI Advertisement by nplusone.ai" with a named sponsor. Where content unlocking is the incentive, the exchange is explicit. A short, qualified survey opens the article, gates fail open on any error, and readers can request a different survey. Incentivized settings can shift who responds; we record the source context on every response so incentivized and non-incentivized evidence can be read separately.
Approved questions, bounded follow-ups.
Opening questions and answer options are approved before a study runs. The AI moderator's follow-ups clarify and probe, inside platform safety rules that constrain what it may say and do. There is no improvising claims, no eliciting personal information, no off-script commitments.
Found against the purpose, cited to the source.
Web research gathers reviews, forums, community threads, and articles against the study's stated purpose. Every extracted signal keeps its source link, and verbatims open the original page. Observed language is reported as observed; it is never blended silently with solicited responses.
Consent first. Redaction at rest.
Consent is resolved before any request through the industry frameworks, and opt-out signals like Global Privacy Control are honored. Personal identifiers (phone, email, card, SSN, IP) are redacted before storage, conversation retention is capped at 365 days, an opt-out is available inside every unit, and research is used in aggregate.
What this method is not.
- Not a probability sample. Reach follows the media plan. For population-level claims, weight accordingly or field alongside panel research.
- Not a metered census of the web. Web research is purpose-scoped gathering, not exhaustive monitoring of every mention.
- Not immune to context effects. Where people answer (mid-article, at a content gate) shapes who answers. We keep the context on the record so you can account for it.
A method that states its limits is a method you can defend in the room. That's the trade we choose.
Interrogate the method before you buy it.
We'd rather answer these questions early. Bring the hard ones.
Talk to us