Consumer Pulse 2026: 74% Would Trust an AI Agent Over Their Best Friend

Consumers now trust an AI agent more than their closest friend

Nearly three in four consumers (74%) would trust a personal AI agent more than their best friend to make a purchase on their behalf. The finding comes from Factor's Consumer Pulse Research 2026, a survey of 25,590 consumers across 16 countries — Australia among them — run online between January 7 and January 22, 2026, with deep dives across 17 categories. Not a chatbot and not a search engine: an agent that acts, decides and buys at the consumer's instruction.

What changed is the question. The 2025 wave, run with 18,214 consumers, found gen AI building an emotional connection and surprising trust in its recommendations. Consumers have stopped deciding whether to trust the technology and started deciding how much of the job to hand it.

The appetite is already specific. Three in four (74%) would instruct an agent to negotiate deals, resolve complaints, renew subscriptions or reorder. One in three (32%) would ask an agent to make a purchase decision inside boundaries they set, validating the choice before approving payment. And 9% would empower an agent to initiate and complete a purchase on its own, before the technology is fully operational. Across those levels, 85% are open to collaboration with an agent in some form.

Discovery has already moved. Among active users — people using gen AI tools at least once a week — gen AI has overtaken the physical store as the number one discovery channel. For an Australian retailer whose differentiation lives in store layout, staff knowledge or shelf position, that is not a scenario to plan against over three years. It has already happened for the most engaged part of the customer base, without a single change to the retailer's own channel mix.

Delegation follows meaning, not difficulty

The most useful idea in the research is that the delegation dial is not a function of how hard a decision feels. Consumers hand off choices that feel like hassle, even complex ones, provided the outcome carries no personal weight. They hold on to choices tied to identity, relationships or self-expression, even when those choices are straightforward, and sometimes because the searching itself is the pleasure. Difficulty predicts almost nothing; meaning predicts almost everything.

Category data makes the pattern concrete, and the ranking holds at every level of autonomy. Recurring services — insurance, telco and utilities — rank first throughout, at 88.4% for collaboration, 78.9% for task execution, 36.1% for delegated decision-making and 12.5% for autonomous purchasing. Routine replenishment follows at 86.9%, 74.4%, 30.8% and 8.6%; discretionary spending at 86.2%, 74.0%, 30.3% and 8.2%; lifestyle at 81.0%, 70.1%, 30.2% and 8.1%; and travel last at 80.2%, 68.8%, 29.5% and 7.0%.

Watch the categories separate as autonomy rises. At collaboration, first and fifth place sit at 88.4% against 80.2% — close enough that one strategy would serve both. At autonomous purchasing the same two sit at 12.5% against 7.0%, close to a two-to-one gap. A readiness assessment built on the collaboration number alone will badly misprice the risk.

The qualitative work — AI-moderated interviews with 50 consumers in five countries, run between April 20 and May 7, 2026 — shows the same line inside one household budget. A Hong Kong participant was comfortable letting an agent buy detergent, bleach or rice, and even book plane and theme park tickets, but insisted on choosing the hotel room, because the view, the position and the feel of it mattered too much. Category sets exposure long before technology does.

Payment is the line consumers will not cross yet

Readiness varies across the journey, and the barrier is precise. The research scores propensity to delegate across eleven stages, from prompting the need to managing loyalty. Payments are the highest hurdle: only 12% are open to agents making purchase decisions autonomously. Where trust does concentrate is telling: consumers are most open to agent autonomy in negotiation and post-purchase, the stages where effort is highest and emotional stakes are lowest. Delegation is flowing to admin, not to authority.

What earns that trust changes as the dial turns. Trust in AI recommendations is a baseline everywhere, but personalisation becomes critical once agents start deciding, and perceived impartiality emerges as a key requirement for fully autonomous purchasing. An agent that cannot show its working, or is suspected of being paid to prefer one option, stalls exactly where the commercial value sits.

That distribution is a design brief rather than a ceiling. The experience that wins in 2026 is not full autonomy; it is an agent that does the comparison, the negotiation and the paperwork, then hands back cleanly at payment with enough context to approve in seconds. Handback done badly — opaque reasoning, no visible trade-offs, no simple override — pushes the consumer back to doing the whole task manually, forfeiting every gain the agent just made.

The line also moves, and possibly fast. Nearly one in three (31%) say experimenting with low-cost, low-risk purchases would make them more comfortable moving toward autonomous agents. Eighty percent would consider delegating at least one daily activity to a personal AI agent. Once data safeguards, configurable permissions, instant override and clear recourse are in place, the research argues full automation could take off faster than expected. Service is the nearest edge of that shift, which Factor's research on customer service examines in an Australian context.

Loyalty is now conditional and agent-enforced

More than half (56%) of all consumers would instruct their AI agent on which brands to consider, which is genuine good news: brand preference still travels into the agent rather than being erased by it. The number beside it is harder. Some 37% of behaviourally loyal consumers would let an agent switch brands for a better fit: more than a third of an apparently loyal base has pre-authorised its own defection.

Behavioural loyalty has always been the fragile kind. Repeat purchase without active preference survives on friction: re-authorising a direct debit, rebuilding a basket, reading competing insurance plans line by line, sitting on hold to close an account. Every one of those costs falls to roughly zero when an agent does the work. What remains is real preference — a far smaller base than most retention dashboards report.

Trust travels in both directions. It flows to the brands an agent chooses and away from the ones it does not, the way a recommendation from a trusted source gains weight on a good outcome and costs the recommender on a bad one. When the front door is an agent yet fulfilment spans several players, the consumer still forms one verdict: the agent makes the promise, and delivery decides whether it is kept. Factor's research on brand in an AI-mediated market examines what still holds preference in place when the agent is choosing.

Agents close the gap between what brands promise and deliver

Influence is already priced in: 71% of all consumers expect gen AI to influence at least half of their spending decisions over the next 12 months, and 43% put budget and value at the top of the list when instructing an agent to decide. Read alone, those two figures look like a margin threat and a race to the bottom.

The rest of the data says otherwise. Some 26% of active gen AI users have already bought a more expensive item because AI increased their confidence in the decision, and the same proportion increased their basket size. Meanwhile 63% want agents to shop for their idealised self, supporting goals like healthier choices, staying on budget or upgrading more intentionally. An agent optimising for the outcome a consumer actually wants will often not land on the cheapest option — it will land on the one that can be shown to work.

What changes is the cost of verification. An agent works continuously, with persistence, precision and perfect recall: it breaks brands into comparable components, tests claims against delivered reality and forgets nothing. Weak differentiation, inflated pricing and poor product fit surface faster than any human shopper could find them. Value built on habit, confusion or the limits of human attention does not merely erode under those conditions; the research argues it is removed from industries entirely, and Factor's work on value migration traces where that margin goes instead.

The decision layer is where competition moves next

Consumers are asking for outcomes that no single company owns. Some 61% want an agent that shops across multiple grocery retailers on their behalf, splitting baskets. Seventy-one percent want one that can plan and book a complete trip across airlines, hotels and activities. Both assume the agent looks across the whole market, which means the comparison happens outside any brand's own environment.

When the agent builds the shortlist and weighs the trade-offs, the moment of choice moves upstream — often before the consumer has engaged with any brand at all. Competition stops being about where a brand shows up and becomes about whether it is visible inside the systems agents use to evaluate and decide. The report calls this the decision layer: standards, defaults, data and verification logic.

Incumbents are not spectators. Consumers are more than 2x as likely to trust brand- or retailer-owned agents as AI-native platform agents, and the top three reasons are existing knowledge of their shopping preferences, trust built through strong service and support, and access to a large selection of products and services. Ascott, a global hospitality and lodging operator, is redesigning its digital infrastructure so its properties are visible inside algorithms, making inventory machine-readable and evolving its AI concierge into an agent that can plan and complete bookings.

Four positions are emerging, and they are not mutually exclusive. A brand can become the consumer's agent of choice, which suits organisations with deep shopper knowledge, first-party data and the ambition to own the decision layer. It can partner with existing agents, trading control for reach. It can build authority through verified claims and clear inventory so agents pick it on its own terms. Or it can serve as a fulfilment partner for demand agents have shaped. The habits consumers form now will be hard to displace — a point developed in Factor's work on platform strategy for agentic AI.

Machine legibility has become a distribution channel

Mondelēz found that Oreo — one of the most recognised brands in the world — appeared in only 10% of AI chatbot recommendations for cookies. The response was not a campaign. The company is overhauling its $3.5 billion digital commerce strategy: unblocking the crawlers that agents rely on, restructuring its digital presence and making product content machine-readable across the portfolio.

Its VP of global digital commerce reduced the problem to one sentence: “If they can’t crawl our site, nothing else matters.” The logic generalises far beyond packaged food. Agents parse structured attributes, cross-check verified claims, weigh price-to-value ratios and score fulfilment track records in milliseconds, with no emotional loyalty bias at all. A brand that has not made its value legible to machines is invisible at the point of decision, however much consumers love it.

This is a parallel discipline to everything marketing already does: agentic engine optimisation alongside search and generative engine optimisation. It means products, services, claims, pricing, availability, policies and proof points that are structured, consistent and machine-verifiable. Brand perception now forms in two places at once — in the consumer's regard and in the agent's evaluation — and both have to hold. Measurement has to move with it: traditional metrics will not tell you whether consumers are being steered toward you or away from you, so the questions become whether you appear in agent recommendations, make the final cut, are named directly and can be selected without friction.

Why this lands hard on Australian organisations

One caveat first. Australia is one of the 16 countries sampled, but the figures above are reported at total-sample level rather than split by market, so they read as direction of travel rather than as an Australian measurement. The Australian implication is analysis — and the report flags retail, travel, financial services, telecommunications and subscription businesses as the sectors where the local lens matters most.

Category concentration is the first reason. Recurring services top the delegation ranking at every stage, reaching 88.4% openness to collaboration and 12.5% openness to fully autonomous purchase. Australian household budgets are weighted heavily toward exactly those lines — energy, insurance, telco, transaction accounts — already the most compared and most switched part of the market. The annual renewal comparison Australians do reluctantly, once a year if at all, is precisely the work an agent performs continuously and unasked.

Market structure is the second. Several concentrated Australian categories have priced on inertia for years, relying on the cost of switching rather than the strength of preference to hold customers. An agent that re-shops a category every month removes that cost without the consumer forming any new opinion, and 37% of behaviourally loyal consumers already say they would allow it. The revenue exposed is not the customers who dislike you; it is the ones who never had a cheap reason to leave.

The third factor argues against retreat. Some 87% of consumers agree AI will impact the role of the store in some way, and 31% say stores will become even more important for creating moments of joy, while 64% of active users say gen AI helps them feel seen, heard and understood. As agents absorb the routine, direct engagements become fewer and more deliberate, raising the value of every human moment an organisation owns. The failure mode is not being disrupted by an agent; it is being unreadable to one.

What Australian marketing and commerce leaders should do next

Audit how your products present to an agent rather than to a shopper. Unblock the crawlers, structure the product data, make claims and pricing machine-verifiable, then check what an agent returns when asked about your category. Oreo surfacing in only 10% of AI chatbot recommendations for cookies is what happens to a category-defining brand that skips this step.

Re-segment loyalty by preference rather than repeat purchase. With 37% of behaviourally loyal consumers willing to let an agent switch them, repeat-purchase rate has stopped being a usable proxy for retention. Model revenue at risk in each category assuming switching friction disappears, and prioritise where the gap between habit and preference is widest.

Design for handback at payment instead of for full autonomy. Only 12% are open to agents making purchase decisions autonomously, while delegation concentrates in negotiation and post-purchase. Build the experience that does the work, then returns a decision a person can approve in seconds, with trade-offs visible and the override immediate.

Use low-risk categories as the proving ground. 31% name a successful low-cost, low-risk purchase as what would move them toward autonomous agents, making replenishment and recurring services the on-ramp rather than high-consideration goods. Earn permission where stakes are low, then extend it.

Decide which seat in the agentic ecosystem you intend to take, and resource it properly. Becoming the agent of choice, partnering with existing agents, building authority so agents choose you and acting as a fulfilment partner each demand different capabilities. That consumers are more than 2x as likely to trust a brand- or retailer-owned agent makes the first option genuinely available to organisations with first-party data. Working through that choice with other Australian executives is usually faster than doing it alone.