Dark patterns: the designs that trick users, and what Australian research found
Pre-ticked add-ons, fake countdown timers, cancellations hidden five screens deep. Australian research found most people have been caught by one.
In short
Dark patterns are design features that steer people into spending, signing up or sharing more than they meant to. The Consumer Policy Research Centre's 2022 survey of 2,000 Australians found 83% had experienced a negative consequence from one, and one in five had spent more than they intended.
A dark pattern is a design that works against the person using it: an option chosen for you, a choice made hard to find, a pressure that isn't real. The term covers everything from a pre-ticked box to a subscription that takes one click to start and a phone call to stop.
What the Australian research found
In June 2022 the Consumer Policy Research Centre published Duped by Design. It surveyed 2,000 Australians in April 2022, a nationally representative sample, and swept a random selection of Australian websites and apps for dark patterns.
- 83% had experienced one or more negative consequences from a website or app using design features to influence their behaviour.
- One in five had spent more than they intended, and almost one in six felt pressured into buying something.
- Younger consumers were 65% more likely to spend more than they intended, and 34% more likely to sign up to something by accident.
The report explored ten types of deceptive design, from hidden costs added to a transaction to scarcity cues creating a sense of urgency that may not be accurate. One example the CPRC gave: an appliance retailer automatically adding a $160 three-year care plan to the cart when someone bought a washing machine.
Common dark patterns, and the honest version
| Dark pattern | What it does | The honest version |
|---|---|---|
| Pre-selected add-ons | Adds extras unless you notice and remove them | Offer the extra; leave the box unticked |
| False urgency | Countdown timers or "only 2 left" that aren't true | Show real stock and real deadlines only |
| Hard to cancel | Easy to sign up, hard to leave | Cancel the same way you joined |
| Confirmshaming | "No thanks, I don't like saving money" | A neutral "No thanks" |
| Hidden costs | Fees appear at the last step | Show the total from the start |
Why it matters for designers
Dark patterns can lift a metric for a quarter and cost trust for years. They also sit close to the line of Australian consumer law on misleading conduct, which is why the CPRC called for new protections. A design review should check for them as routinely as it checks contrast.
Persuasion, usability mistakes and manipulation
A prominent purchase button is not enough to establish a dark pattern. A shop needs to help people complete a purchase, and a subscription service needs to explain its paid plans. The question is whether the interface gives people an accurate understanding and a practical choice.
An awkward form might be an ordinary usability defect. If it rejects a valid postcode, the likely remedy is to repair validation. A cancellation form that accepts ordinary text until someone asks to cancel, then repeatedly rejects that request without an explanation, needs a wider investigation: the behaviour might systematically obstruct a choice the business dislikes.
Intent is often hard to establish from a screenshot. Record the observable problem first. “The optional paid extra is selected when the basket opens” is more useful than “the designer wanted to deceive everyone”. It identifies something another person can reproduce, gives the product team a concrete repair and avoids pretending to know a stranger's motives.
Likewise, a genuine deadline is useful information. The issue with a countdown is whether its meaning is true. Does the offer actually finish when it reaches zero? Does the timer restart on a new visit? Does the stated limitation apply to the product being considered? Test those questions before describing the design as false urgency.
Worked example: an optional delivery extra
The following is an invented interface example, not a finding about a named retailer. Imagine a shop selling a desk lamp for $60, with ordinary delivery costing $8 and optional delivery protection costing $4. These figures are illustrative.
In the problematic version, the product page promotes the $60 figure, the basket silently adds protection, and the checkout button says “Continue”. A customer reaches a payment screen showing $72 without a clear explanation of the extra. Someone who notices the difference has to expand a small panel to remove the protection.
A review would record the exact starting conditions: a new basket, one lamp, ordinary delivery and no previous choice about extras. Screenshots would show the basket before and after opening the hidden panel. The reviewer would calculate the displayed total from the items actually selected and check whether the same behaviour occurs on a phone and with a keyboard.
The repair is not simply to make the total font bigger. Present the $4 protection as an optional, initially unselected choice, explain its actual benefit and show the delivery charge and final total before payment. The customer can then choose a $68 order or a $72 order with the extra, understanding the difference.
A useful acceptance check is: a new customer who takes no action on optional protection reaches payment with protection absent. Selecting it adds exactly $4, removing it subtracts exactly $4, and both states name the extra beside the total. This proves the behaviour changed; it does not rely on a team declaring that the screen now feels more transparent.
Review the whole subscription journey
A subscription can be clear on its pricing page and confusing later. Walk through joining, the trial period, renewal, changing a plan and leaving. Each stage can introduce a different mismatch between what someone expects and what happens.
For an invented trial, write down what the person sees before confirming: the trial duration, the amount due today, what happens after the trial, the recurring price, billing frequency and where to cancel. Put those facts near the commitment. A link to lengthy terms should support the explanation rather than carry the entire explanation on its own.
Check the confirmation too. If the interface says “Trial started”, the account should identify when paid billing begins and how to manage it. A customer who returns later should not need to recreate the original signup journey to learn their own plan terms.
For cancellation, distinguish four outcomes: stopping renewal, ending access immediately, pausing billing and deleting the account. These are different actions. Calling all four “leave” can create accidental outcomes even without deceptive intent. Use the actual action and explain the effect before confirmation.
A retention offer can be presented without replacing the requested action. For example, an optional pause offer can sit beside a clear way to continue cancellation. After cancellation, show a receipt that names the effective date and any remaining access. Avoid a success-looking screen when the request has merely been saved for someone else to consider.
These are design review questions, not a statement that one cancellation process is legally required in every Australian contract. Obligations depend on the transaction and applicable law. Where the product team needs a legal conclusion, retain the actual screens, terms and dates for a qualified review.
Consent and personal information
Consent screens deserve the same attention as payment screens. Identify which processing is necessary for the requested service, which choices are optional and what happens if the optional choices are declined. Do not describe a data use as optional if the product refuses the core service when someone declines it without explaining why.
In an invented newsletter form, the task is to receive a monthly product update. Adding unrelated partner offers through a preselected control changes that task. Separate the choices, name the sender and explain the expected content. A person should not have to interpret a double negative to avoid a second mailing list.
Test reversal as well as the initial choice. If someone changes a preference, does the account show the new state after a reload? Does a confirmation email describe what actually changed? Are the same choices available when using a screen reader or when text is enlarged?
Do not use a visual review as a privacy compliance certificate. Interface wording cannot establish what a server stores or which third parties receive it. Pair the visible review with the product's actual data flow and ask the relevant privacy specialist to resolve obligations that the screen alone cannot answer.
A practical dark-pattern review method
Start with one real task rather than a list of screenshots. Suitable tasks include buying one item without extras, declining marketing, finding the full recurring price, cancelling renewal and comparing two plans. Use a safe test account and avoid submitting a real payment merely to gather a screenshot.
Record the starting state: device width, account status, location or currency selection, existing subscription and any saved preferences. A screen may behave differently for a returning member and a new visitor. Without these conditions, another reviewer may be unable to reproduce the finding.
Move through the journey in order and capture the moments that affect the decision. For each one, note the claim, the available actions, the information visible before choosing and the resulting account or basket state. Open relevant explanations and record whether they clarify the main message or contradict it.
Then try the less profitable choice. Decline the extra, select the lower plan or continue cancellation. Count the required steps, but also describe them: an extra confirmation that prevents account deletion is different from an unrelated questionnaire that blocks cancellation. A step count without context can mislead the reviewing team.
Finally, repeat the critical path with keyboard navigation and at a phone width. A visible decline link is not a meaningful choice for someone who cannot reach or activate it. Conversely, a hidden element in the source is not proof that the rendered interface offers it. Verify the behaviour a person actually encounters.
Write findings that can be fixed
Use a short record for each finding. Include the task, starting conditions, steps, observed behaviour, likely consequence and proposed repair. Keep the observation separate from an assumption about why it happened.
For example: “In the annual plan selector, the large price is shown per month while the payment confirmation charges a full year. The annual commitment appears only after opening an explanation.” The potential consequence is that a customer expects a smaller immediate charge. The proposed repair is to identify the billing commitment and total due next to the price before selection.
Prioritise by the consequence and the reach of the problem. An unwanted charge, lost access or disclosure of personal information may need urgent containment. A confusing secondary label may be less urgent, although repeated confusion still deserves repair. Record what is known about frequency; do not turn a single reproduced case into a claim about every customer.
Give the repair an observable completion condition. “Improve trust” is an aim, not a test. “The annual price card states the full annual charge before the customer selects it, and the confirmation shows the same charge” can be checked directly.
Retain before-and-after evidence privately. Use fictional details in a shared design example, and remove personal information from screenshots before circulating them. The point is to preserve enough context for verification without exposing a customer's purchase or account history.
Measure outcomes without rewarding confusion
A purchase rate alone cannot explain whether people understood the purchase. If a redesign adds paid extras more often, investigate whether customers intentionally selected them, whether refunds changed and whether support received more “I didn't choose this” complaints. Do not count an accidental purchase as a successful experience simply because the payment completed.
For an experiment, state the intended benefit in advance. One possible hypothesis is that clearer billing language reduces mistaken annual purchases while preserving informed signups. Useful evidence might include comprehension during moderated tests, billing-related support themes and successful completion of cancellation. The appropriate measures depend on the actual product and its data access.
Keep observations proportionate. Five test participants misunderstanding a label demonstrates a problem worth investigating; it does not establish the percentage of the entire customer base affected. Similarly, a lower complaint count may reflect fewer customers or a harder-to-find contact route. Check the denominator and the way the evidence was collected.
An honest interface may reduce a short-term metric that previously included unwanted actions. That is a reason to redefine the desired outcome, not to hide the decline or unsubscribe option again. Record the commercial effect alongside customer understanding so the business can evaluate the change with the right context.
Australian research and current claims
The CPRC figures above describe research published in 2022, not a fresh measurement of all Australians in 2026. They are evidence about the study's participants and methodology. They should not be presented as this site's research, or as a live prevalence counter.
For current Australian consumer guidance, the ACCC's false or misleading claims page explains that businesses need accurate information and that qualifications must not contradict an advertisement's overall impression. That is a useful reason to review the prominent promise and the nearby explanation together, rather than treating fine print as an automatic cure.
This guide does not label every frustrating interface unlawful. It helps a designer identify observable choices and preserve evidence. A specific legal assessment needs the actual conduct and circumstances, including what was represented and what the customer could reasonably understand.
Where AI review helps, and where people are needed
An AI review can suggest places to inspect, such as a missing recurring price or an unclear button label. It cannot establish from one screenshot whether a timer is genuine, a cancellation request reaches the billing system or a preference is honoured. Those questions require interaction and, sometimes, internal records.
Use usability heuristics to structure an expert review, user testing to observe how suitable participants understand a journey and the contrast checker to examine colour pairs. These methods answer different questions. A contrast pass does not make a misleading offer fair, and a clear offer still needs readable controls.
Ideal User Interfaces' proposed AI reviewer remains in design. The review method and invented examples here are available to use today; they are not results from an operational AI audit service.