You are comparing two apps at the kitchen table. Both say “science-backed”. One links to a paper; the other shows a cheerful diagram of the brain. Neither has yet explained what its claim means for your next difficult conversation.
A science-backed relationship app should let you inspect the research behind its particular promise. Some digital programmes have encouraging evidence. That evidence belongs to the intervention, people and outcomes studied. Before paying, check whether the paper tested the app you are considering and what its design can tell you.
Begin with the promise, then open the paper
Write down the claim in ordinary language: “This app helps us communicate”, for example. Look for the original research rather than stopping at an expert photograph, review score or a company’s summary.
Ask what was studied. Was it a theory about relationships, a classroom course, a guided online programme or the actual app? Those can all be useful sources of knowledge, but they support different claims. A programme drawing on a therapy approach is not automatically therapy, and a theory-informed feature has not necessarily been evaluated.
You do not need to become a statistician. Start with the question the researchers asked and the comparison they made. If you cannot find those, ask the provider for the study behind its exact claim.
Read an app evaluation in its own context
A 2025 mixed-methods evaluation of Paired used surveys and interviews with self-selecting users. It found encouraging associations between use and relationship quality. It was not a randomised trial comparing the app with another intervention or no app.
That distinction matters: a positive association cannot by itself establish how much change the app caused. The researchers describe the work as proof of concept and discuss sampling limits. The paper also discloses an author’s unpaid role at Paired and describes independent analysis.
Use that study as evidence about an early version of Paired, rather than a guaranteed improvement for a new user or proof about a different product. A percentage in an advert needs the study’s actual comparison beside it.
Source: Paired mixed-methods evaluation, JMIR mHealth and uHealth.
Notice when a programme has a comparison group
A 2016 OurRelationship trial randomly assigned 300 heterosexual US couples to a defined online programme or a two-month waiting list. The programme included online activities and four brief staff calls. Couples in the programme reported greater improvements in relationship satisfaction than the waiting-list group.
This provides evidence for that particular intervention and comparison. It does not establish that every self-guided couples app works, or that an app is equivalent to individualised therapy. The study excluded some higher-risk situations, including recent partner violence involving injury or fear. Its authors also disclosed a financial interest in the programme’s intellectual property.
Read support level, eligibility and comparison together. A guided programme and a daily question widget can ask very different things of a couple.
Source: OurRelationship randomised controlled trial.
Check what the result actually measures
More completed prompts show engagement. They do not, on their own, show that partners understand each other better or that harmful behaviour has stopped. Look for the outcome you care about and how it was measured.
If a study uses self-reported relationship quality, describe its result that way. Do not quietly turn it into observed conflict reduction, a clinical treatment effect or a lower chance of separation. A result measured after a short period also cannot tell you everything about the following year.
Check who stayed in the study. People who continue using an app and completing surveys may differ from those who leave. Ask how missing responses, follow-up and the product version were handled. Funding and researcher relationships belong in that picture too; disclosure helps you interpret a study rather than automatically accepting or dismissing it.
Turn the claim into a small decision
Before choosing a subscription, make a short evidence note. Use it to organise what you know and what you still need to ask.
1. The promise: what does the provider say will change?
2. The evidence: which named intervention, people and outcome were studied?
3. The limit: what important question does that design leave unanswered?
4. Our next step: is there a modest use we both welcome, or do we need different support?
Suppose you want to talk about evenings that disappear into separate screens. Try one ordinary question when both people agree: “When would you like a little company this week?” Listen to the answer and agree something manageable. Later, ask whether the conversation became clearer and whether the agreement happened.
That personal review helps you judge fit. It is not a clinical trial, and one pleasant evening cannot establish the product’s effectiveness. A tool can still be useful without claiming more than you have observed.
Keep evidence, consent and appropriate care together
Even an encouraging study cannot make a particular sharing arrangement acceptable. Read current privacy and subscription terms, check who sees answers, and make sure both people can pass or stop. Consider whether using the tool gives one person another responsibility to organise the other’s participation.
Persistent distress or clinical concerns deserve appropriately qualified assessment. An app exercise should not delay it. Where there is coercion, abuse or fear of retaliation, seek confidential specialist support rather than using joint questions to obtain disclosures or settle a dispute.
Back at the kitchen table, the useful choice may be a small trial, a question to the provider or a different kind of help. “Science-backed” becomes more useful when you can finish the sentence: backed for what, for whom, and with which limits?

