Me, My Brand and AI: When 36% of Users Call Generative AI a Good Friend

Generative AI has stopped being a tool and become a confidant

36% of active generative AI users consider the technology a “good friend”. That is not a finding about capability but about relationship — and relationship decides whose recommendations a person acts on. A tool gets used. A friend gets believed. The distance between those two verbs is why this research belongs on a marketing agenda rather than a technology one.

Adoption has already crossed into the mainstream, with 72% of consumers using generative AI tools regularly. Reach at that level puts the technology alongside the largest media channels a brand buys. What makes it different is not the size of the audience but the nature of what people bring to it.

94% of active generative AI users have asked, or would consider asking, for help with personal development goals, and 87% say the same for social and relationship advice. Those are not shopping queries. They are the conversations people reserve for a partner, a close friend or a paid professional, and a system occupying that seat is not judged the way a search box is. Confiding in something changes the standing of whatever it says back.

The research sorts the shift into three roles: the trusted guide whose suggestions shape what people consider, the loyal companion that puts a person's needs first, and the second self trusted to act on their behalf. Each role moves AI further inside the decision — informing it, then shaping it, then making it — and progress in two areas of cognition, empathy and autonomy, carries it from one to the next. Most marketing plans are written only for the first.

One in ten consumers already trusts AI above every other source

Nearly one in 10 consumers (9%) already rank generative AI as their single-most trusted source of what to buy. The precision matters: not one trusted source among several, but the most trusted one — ahead of friends, reviews, retailers and the brands themselves. For a channel most media plans still classify as experimental, that is a remarkable position.

Behaviour is running ahead of stated trust. One in two users have already informed a purchase decision using generative AI, making it the fastest-growing source for recommendations and advice in the past year. Half of users consulting it while only 9% will name it their most trusted source is the same pattern that preceded search engines and online reviews becoming default infrastructure.

The growth rate, not the level, is the number to plan against. A 9% top-trust share is roughly what a serious challenger brand spends years fighting to win, and here it has accrued to a channel no brand owns or can buy its way into. The question is not whether 9% justifies attention today, but what the same trajectory produces by the end of a three-year plan.

Generative AI is already the second place consumers go for a recommendation

Among active generative AI users, the ranking of preferred sources for purchase recommendations now puts physical stores first at 19% and generative AI second at 18%. A single percentage point separates a channel built over more than a century from one built in a handful of years. Social media and online marketplaces follow at 15% each, friends and family at 13%, and search engines at 11%.

The figure that should trouble brand owners sits at the bottom. A brand's own website or app is the least-preferred source of the seven, at 10% — behind the search engine, behind the marketplace, behind the friend. The owned channel most organisations have spent a decade optimising is where these consumers are least inclined to go for advice on what to buy, which recasts much digital investment as conversion infrastructure rather than influence.

The mechanism is intermediation. As third-party large language models expand their influence from discovery through to conversion, a brand that does nothing risks being misrepresented or dropped from the consideration set altogether. The model does not read your homepage before it answers; it answers from whatever representation of your brand it has assembled out of content, reviews and third-party signals. In most organisations that representation is a material marketing asset with no owner.

That makes the LLM ecosystem — the network of models, platforms, data sources and integrations shaping how AI tools respond to consumer needs — a distribution channel needing its own strategy, assessed platform by platform and category by category. Direct partnerships let a brand shape the experience on its own terms rather than being passively represented. Discoverability is the pathway no one can skip: high-quality, frequently updated content and credible third-party signals such as reviews and social posts. Factor's work on platform strategy in the agentic era sets out how those choices interact.

Uncertainty is the emotional weather, and it pushes categories towards price

More than half (54%) of consumers now see uncertainty as the new normal, a sentiment that has doubled in the past year. The doubling matters more than the level. A stable majority would be a condition to design around; a figure that has doubled in twelve months is a fast-moving change in how people approach every discretionary decision they make.

Two forces now point the same way. Diminishing consumer confidence and agentic solutions together threaten a race to the bottom in which brands compete on price. An anxious consumer narrows to what feels safe and cheap, while an agent optimises on the criteria it can read — and price and specifications are the easiest to read. If agents optimise only on those, brands become interchangeable by default rather than by choice.

The answer this research points to is not louder messaging but differentiation on experience and connection, in ways relevant to the category and the individual consumer. That is a harder brief than a price promotion and a slower one to execute, which is exactly why it holds its value: a discount is matched in a week, an experience is not. Readers tracking how consumer sentiment is moving will find the companion picture in Factor's Consumer Pulse research.

Emotional experience is the one thing an algorithm cannot commoditise

More than one-third (34%) of consumers would switch from a preferred brand to one that makes them feel special. Preference, in other words, is not loyalty. A third of the people who already choose you are one better experience away from choosing someone else, and in a time of heightened economic uncertainty that decision gets made faster and with less guilt.

The upside is measured in the same currency. Consumers are 1.5x more engaged, 2.3x more likely to recommend and 1.7x more likely to accept a higher price point from a brand that delivers emotionally engaging experiences. The price multiple is the one to carry into a budget conversation: emotional engagement is not a brand-health metric that pays off eventually, it is margin protection visible in what people will tolerate paying this quarter.

Loyalty programmes are the most efficient place to start. Loyalty programme members are 1.6x more likely to be experientially or emotionally motivated, more willing to share data and engage with personalised experiences, and twice as likely to help brands refine new products and services.

Brands are better placed to build this than the platforms are. A brand's comprehensive understanding of its own consumers and its category is what allows an experience to be tailored, proactive, personal and empathetic at once. Consumer demand is running towards depth, authenticity and sensory richness, which favours physical experience combined with multi-modal methods — augmented reality, video, voice and images — over another personalised email. The platforms have the scale; the brand has the context, and context is what makes an experience feel meant for one person.

Proactivity is the brand behaviour that survives translation into an agent

Proactivity makes brands 19% more likely to be preferred by consumers. The shift it describes is from responding to anticipating — flagging a fault before the customer notices it, resolving a billing problem without being asked, arriving with the answer before the question has formed. AI makes that affordable at scale, because prediction engines can now simulate consumer behaviour well enough to tell a brand where to act first.

Proactivity is also the one brand behaviour that translates cleanly into an agent-mediated journey. An agent will not relay your tone of voice, your art direction or your sponsorship of a code. It will register that you acted first, because acting first produces an event — a fault avoided, a cost saved, a delivery rescheduled — and events are exactly the kind of signal a machine can read, store and score.

There is a counterweight, and it is sharp. Consumers distrust AI-generated content lacking authenticity (41%) and personability (45%). Close to half the market penalising output that feels impersonal is a direct caution against the obvious efficiency play, because the technology that makes proactivity cheap also makes flat, generic content cheap. Volume without a defined personality does not merely fail to land; it subtracts trust the brand had already earned.

The practical standard is therefore higher, not lower. An AI-mediated interaction has to carry an authentic personality and feel natural and intuitive, which means deciding what the organisation sounds like before automating what it says. One dependency is routinely missed alongside it: as generative AI shifts into commerce, a personalised recommendation only helps if the item is available, making alignment with supply chain data a precondition.

The agent is becoming the buyer, and it does not see your advertising

75% of consumers are open to using a trusted AI-powered personal shopper that understands their needs. Three in four is not an early-adopter fringe; it is latent demand waiting for a product good enough to claim it. Intelligent agents can already act on instructions and buy on a consumer's behalf, which moves AI from influencing the decision to making it.

The commercial exposure sits in the media plan. As bot-to-bot commerce becomes more integrated, traditional brand touchpoints such as banner ads and other retail media may be bypassed altogether. Product comparison, checkout and post-purchase support get streamlined by an intermediary that never renders an impression, so budget attached to human impressions is the first line to lose its footing.

Existing search equity is not wasted. Search engine optimisation still matters to language models, and generative engine optimisation can drive organic visibility in AI-generated results; Factor's work with a global retail client found that top-ranked traditional search results also rank well with LLMs. What changes is the objective: ranking for a query becomes a means to being cited in an answer, and being cited becomes a means to being available to an agent comparing options.

Data is what gives a brand negotiating power in that ecosystem. Consolidating zero-party, first-party and contextual data — location, time, device, weather and, crucially, past interactions — into intelligent consumer profiles is what lets AI engage in ways that feel genuine rather than generic. Commerce assets carry the same weight: real-time inventory, pricing and location-specific information enable frictionless transactions and are a bargaining chip when partnership terms are set. Trust has to be engineered alongside them, through data protection, consent-based personalisation, cybersecurity and transparency about how AI-powered interactions are shaped. Factor examines the transaction end of the same shift in the dawn of the agentic deal.

What this means for Australian organisations

One point should be stated plainly before any of this is applied locally. The Consumer Pulse Research 2025 base captured responses from 18,000 consumers in 14 countries, and this Australia edition retains that research rather than re-collecting it: Factor has localised the framing, not the sample. There is no separate Australian break-out, so what follows is analysis of what globally observed behaviour implies for Australian organisations, not a measurement of Australian consumers.

The structure of the Australian market sharpens the risk. Grocery, banking, telecommunications, insurance and domestic aviation are concentrated categories in which consumers choose among a small number of large brands offering broadly comparable propositions. Narrow assortments are exactly the conditions under which an agent reaches a price-and-specification answer quickly, so the race to the bottom this research warns about would arrive faster here than in fragmented markets. Differentiation on experience is not a premium play here; it is what stands between a category and commodity pricing.

One finding cuts in Australia's favour. Physical stores remain the highest-ranked source for purchase recommendations at 19%, and Australian retailers, banks and insurers still run dense branch and store networks that many international competitors have dismantled. That is the one recommendation channel an agent cannot enter — yet it is usually managed as a cost line rather than as the highest-ranked recommendation source in the data.

There is also a representation problem specific to a smaller market. A model's answer about your brand is assembled from the content and third-party signals available to it, and an Australian brand with a domestic-only content footprint generates far less of that material than a global competitor selling into the same local category. The remedy is this research's own guidance applied with local discipline: frequently updated quality content, credible third-party signals, and trust practices treated as marketing infrastructure, not compliance overhead.

The loyalty finding lands differently here too, where supermarket, airline and retail schemes enrol a large share of the population. Many Australian organisations already hold the member base this research identifies as 1.6x more likely to be experientially or emotionally motivated. Most run those schemes as discount mechanics; run as a co-creation channel instead, they are the fastest route to the experiential differentiation this research argues for. Explore the Factor research library or join a Factor event to work through these implications with other Australian brand and marketing leaders.

What Australian marketers should do next

Audit how the major models describe your brand and your category before buying another impression. With 9% of consumers ranking generative AI their single-most trusted source of what to buy, and generative AI second at 18% among active users' preferred recommendation sources, the model's answer is a shelf position you do not own. Give it an owner, a baseline and a monthly measurement, as you would share of search.

Fund proactivity ahead of personalised messaging. Proactivity carries a measured 19% preference lift and is the one brand behaviour that survives translation into an agent-mediated journey, where tone and creative do not. Name the three moments where you could act before the customer asks, confirm your supply chain data supports acting on them, and build those before the next campaign.

Treat emotional experience as a margin decision rather than a brand one. Emotionally engaging experiences make consumers 1.7x more likely to accept a higher price point, 1.5x more engaged and 2.3x more likely to recommend — the strongest defence against agents that would otherwise sort your category on price. Take those multiples into the budget conversation instead of a brand-tracker slide.

Start with the loyalty base, and change what you ask of it. Members are 1.6x more likely to be experientially or emotionally motivated and twice as likely to help refine new products and services — both the fastest test bed for AI-led experiences and the most willing source of the first-party data they need. Co-create with that group rather than surveying it.

Set an authenticity standard before scaling AI-generated content. Consumers distrust AI-generated content lacking personability (45%) and authenticity (41%), so volume without a defined brand personality subtracts trust. Write the personality specification for your automated interactions, review live output against it, and fix the voice before increasing throughput.