I asked Woolies’ AI shopping assistant to plan my weekly shop. Here’s what happened.

This week I finally had the chance to test drive the new Olive assistant experience in beta and get under the hood of some of the new features in the Woolies app.

In late June, Woolies announced that the beloved Olive - who I’ve been a fan of for a long time - was getting a glow-up through Woolies’ partnership with Google.

The new Olive promises to personalise meal planning and shopping based on individual preferences, while also helping shoppers manage their budgets by identifying specials and suggesting “smart swaps” to save money.

It was that last feature that intrigued me most. Because while smart swaps might be great news for shoppers, they raise an interesting question for brands: what happens when AI actively encourages your customer to swap your product for a cheaper alternative or private-label equivalent?

At the same time, Woolies introduced two other features to make building the weekly shop easier.

Snap & Shop allows shoppers to take a photo of an item or handwritten shopping list and automatically add the products to their shop (I’d genuinely love to know how many people are handwriting shopping lists only to upload them to an app - perhaps more than I think.)

Then there’s Smart Basket, which predicts recurring purchases to help shoppers build their basket faster and suggests other items they may want to add.

So, with a shopping list in hand, I put the new experience through its paces. Here’s what I found.

The new Smart Basket feature feels like a logical next step. It essentially flips the traditional “buy again” experience on its head. Rather than shoppers going into past purchases and manually adding individual products to their basket, Smart Basket predicts a selection of regularly purchased items and adds them all with the click of a button. Shoppers can then simply remove anything they don’t need.

It’s a smart move by Woolworths and taps into the growing shift towards invisible commerce. For some shoppers, building the weekly basket becomes increasingly automatic - products are added with little active consideration or decision-making.

For brands, that has an important implication. If shoppers increasingly start with a pre-populated basket rather than an empty one, getting into that basket in the first place becomes even more vital - and winning shoppers who are new to your brand becomes even more critical.

Uploading a physical shopping list was one experience that really impressed me. Given shopping lists tend to be generic - “milk”, “bread”, “toothpaste” - rather than specifying exact brands, variants and sizes, I was interested to see what the app would recommend.

For me, the assistant had to pass two tests. First, could it accurately read and interpret my messy handwriting? And second, could it use my previous shopping behaviour to identify the products I was most likely looking for without needing more information from me? If I had to clarify brands, sizes or variants along the way, it would fail the speed test.

It delivered on both.

The app interpreted my handwriting perfectly and recommended products I’d previously purchased. It also surfaced alternatives that I’d bought before, giving me relevant choices without forcing me to start from scratch.

It’s a relatively simple use case, but one that genuinely made building a basket faster.

Olive’s new capabilities are still in beta, so we should expect the experience to evolve. But with Woolies promising more personalised meal planning and budget support, I wanted to see how far the assistant could already go.

I started by asking Olive to suggest products for a vegetarian lasagne. She returned five recommended products and, interestingly, most were branded rather than private-label alternatives. The recommendations also reflected brands I regularly purchase.

That was a big first tick for me. I’m predominantly a branded-product shopper and rarely buy private label for most products, so the recommendations felt aligned to my existing preferences rather than simply defaulting to the cheapest option.

There was another important takeaway for brands, though: for each key ingredient, Olive generally surfaced one primary recommendation rather than giving me a long list of options. 

Next, I asked Olive to suggest meals for a family of four for the week. It returned a number of fairly generic meal ideas and associated products, but what interested me was what it didn’t do.

Olive didn’t ask me any questions before making its initial recommendations - about dietary preferences, budget, how many nights I wanted to cook, how much time I had, or even the kinds of meals my family tends to enjoy. So my assumption therefore was, it understands me and will use what it knows about me to make recommendations - but it admitted it didn’t know my past purchase behaviour - which I share more about in a moment.

That therefore feels like an important opportunity. If AI is going to genuinely improve inspiration and meal planning, the quality of the experience will increasingly depend on how well it can combine what it already knows about the shopper with the right questions at the right time. Speed is important but not at the expense of relevancy. 

The bigger challenge: one assistant, multiple shopping experiences

What I’ve been thinking about more broadly is how retailers like Woolies are building shopping assistants alongside existing eCommerce journeys and search and discovery features.

At the moment, these experiences can still feel quite separate.

Olive sits as a distinct tool that shoppers need to consciously open and engage with, rather than operating natively across the broader shopping experience. Search, Smart Basket and Olive each do useful things, but they don’t yet feel like one connected experience or leveraging the same intelligence layer.

I saw that first-hand what happens when the experiences are not connected. When I asked Olive to add products I’d previously purchased to my basket, she told me it couldn’t access my purchase history.

Yet elsewhere in the app, Smart Basket is already using previous shopping behaviour to predict products I’m likely to need.

That creates an interesting disconnect. The capability exists within the Woolies ecosystem, but the customer still needs to know which feature to use to access it.

The more powerful future experience is one where the shopper doesn’t need to understand the architecture underneath it. I shouldn’t need to know that Smart Basket handles repeat purchases while Olive handles meal inspiration. I should simply be able to say: “Sort out my weekly shop” - and have the experience draw on my purchase history, preferences, budget, current specials and meal needs to do the heavy lifting.

A few developments in AI discovery have caught my attention over the past couple of weeks. The pace of change is accelerating offering a glimpse into how much change we can expect to see over the next 12–18 months, and importantly what brands should be paying attention to now.

1. There are two battles for AI visibility: being trusted and being chosen

As brands start measuring their visibility in AI, citations have quickly become a key metric. But new Semrush research suggests they may only tell us half the story.

Semrush analysed 126 million US retail prompts across ChatGPT, Gemini, Google AI Mode and AI Overviews, comparing the brands AI mentions (a brand named in an answer as a recommendation) with the sources it cites (a domain linked beneath it as evidence).

Alibaba was ChatGPT's most-cited retail source, with 4.2 million citations between January and April - yet rarely appeared as a recommendation or mention. Costco sits at the other end of the spectrum. AI frequently recommends or mentions Costco - it was mentioned 1.3 million times - but when explaining or supporting those recommendations with a citation, it often doesn't cite Costco's own website. Instead, AI turns to review sites, forums and even competing retailers as its citation source. In other words, Costco gets the recommendation, but other sources are helping shape the story AI tells about it.

It creates two very different battles for brands: Can you become a brand AI recommends? And can you become a source AI trusts and cites?

Winning the first gets you into the consideration set. Winning the second gives you greater influence over the information AI uses to understand your brand and category.

Ideally, you want both. Because a growing citation count might look like success - but if AI is learning from you while recommending someone else, who is really winning?

2. What happens when AI changes who it trusts?

Something unusual happened in ChatGPT this month. Reddit citations dropped by more than 80%, almost overnight. YouTube fell 88%, TikTok 72%, Facebook 38% and LinkedIn 36%. Yet despite the dramatic shift in sources, the brands ChatGPT recommended barely moved.

That's particularly interesting given the attention Reddit has attracted as brands look to improve their visibility in AI-generated answers.

Petra Labs' research found that much of the citation share lost by social platforms shifted to what it calls "non-competitive owned media" - content published by companies relevant to the topic but not directly competing with the brand being recommended. For example, Zapier might publish a comparison of Google Sheets and Microsoft Excel because its product integrates with both. It doesn't compete with either, but its content can still become a source ChatGPT trusts and cites.

Reddit has since shown signs of rebounding, and the same dramatic movement hasn't been observed across other AI platforms.

So, what does it tell us? Firstly, the citations we see may not tell us the full story of what influences a recommendation. Secondly, today's dominant sources may not be tomorrows.

Reddit still matters. So do YouTube, publishers, creators, reviews and owned content. But this reinforces the importance of building authority across the broader ecosystem rather than relying too heavily on any one source.

3. AI content is getting its own provenance layer

Anthropic has announced that future Claude models will generate text containing an invisible watermark designed to indicate the likelihood Claude was involved in creating it. Supported images and files will also carry C2PA Content Credentials recording that AI was involved in creating or processing the asset. OpenAI is taking a similar approach to generated media, with supported images created through ChatGPT carrying both C2PA credentials and invisible SynthID watermarks.

The catalyst is regulation. From August 2026, new transparency requirements under the EU AI Act have begun applying to generative AI providers in Europe, requiring AI-generated content to be identifiable. Anthropic says it will apply its watermark globally.

We're already seeing this play out across commerce and content platforms, with Amazon requiring disclosure for certain AI-generated imagery and social platforms increasingly labelling synthetic content.

But here's where it gets particularly interesting for brands. Could AI provenance eventually influence discoverability too?

A new study from First Page Sage tracked 1,682 pieces of content across 139 websites and found watermarked AI-created content ranked, on average, five positions lower on Google and was cited by AI platforms 7% of the time versus 12% for un-watermarked content. 

There are important caveats: the study shows correlation, not causation, and didn't fully control for differences in content quality. But it's one worth watching.

As AI becomes embedded into product imagery, descriptions, advertising and other commerce content, content provenance may not just become another layer of digital shelf governance - it could potentially become another signal influencing visibility too.

Marketplaces have evolved from the fringes of retail to one of its most significant growth engines. With nearly one in three Australian retailers now operating a marketplace, brands are rethinking how and where they compete.

In this episode, Teresa is joined by Mark Mansour, Managing Director of Woolworths MarketPlus, to explore the rapidly changing marketplace landscape. Together, they discuss what's driving retailers to launch marketplaces, how brands should think about where to play and how to win, what strong retailer-brand partnerships look like, and how AI and agentic commerce may shape the future of marketplaces. A must-listen for brands navigating the next era of digital commerce.

📊 Retail media’s fragmentation problem is getting harder to ignore. As brands invest across a growing mix of retail media networks, marketplaces and quick-commerce platforms, the challenge is increasingly becoming how to bring the data together. Flywheel estimates some FMCG teams are spending up to 60% of their week pulling fragmented retail media data together, leaving less time to actually optimise performance. As retail media investment continues to grow, the winners may not simply be those spending more, but those that can build the data, measurement and operating capability to understand what’s working across the ecosystem.

✈️ Velocity and CommBank are bringing loyalty closer to everyday banking. From October, CommBank’s Yello program will see more than nine million CommBank customers be rewarded for leveraging an array of services across the bank including home loans, insurance and credit cards - meaning for the first time customers will be rewarded for their everyday banking not just for their credit card purchases. The reimagined CommBank Yello will deliver enhanced banking rewards, working with established brands including Woolworths Everyday Rewards, Origin Energy, Myer, Velocity Frequent Flyer, DoorDash, BP, and Qatar Airways making rewards clearer, more useful and easier to use. This is a major step by an Australian bank, which will no doubt likely see others follow in an already cluttered loyalty landscape.

🛒 Retailer update: eCommerce growth and faster fulfilment are raising the bar.

Coles’ supermarket eCommerce sales jumped 26.4% to $5.6 billion in FY26, reaching 13.6% of supermarket sales. Importantly, its automated Customer Fulfilment Centres turned EBITDA positive in their second full year of operation, with CEO Leah Weckert saying the eCommerce business is now scaling profitably, not just growing.

Meanwhile, BIG W has launched same-day delivery nationwide through DoorDash, with thousands of products available for delivery by 5pm when ordered before midday. BigW is the latest in a string of retailers partnering with quick commerce platforms to embed quick commerce as a core part of their omni-channel strategy.

I’m an advisor, trainer and thought leader with 25+ years’ experience across digital, marketing, loyalty and commerce, working with consumer, FMCG and retail brands. I’m also the host of Unpacking the Digital Shelf – APAC Edition and have been recognised as a RETHINK Retail Top Retail Expert (2026), and Top 20 CMO (2018).

Before you go…

🧭 Need strategic clarity or expert advice? Let’s chat.

🛠️ Looking to build capability in your team? Explore training & workshops.

🎤 Planning an event or offsite? Book a keynote or thought leadership session.

Get in touch at [email protected] or find out more here, and don’t forget to follow along on LinkedIn and Substack.