Product discovery and AI: Part I
Tech continues to shape how we shop - here's the latest
In The Devil Wears Prada, Miranda Priestly, played by Meryl Streep, has an iconic monologue about the power of the fashion industry - exemplified by the cerulean sweater. She shares how a creative design decision made at a luxury fashion house - the decision to use a certain shade of blue - trickles down to the high street, where, without even choosing to, the average consumer participates in these trends.
I think a lot about how what we buy is shaped not only by creative decisions, but by financial ones. Namely, how brands spend money to ensure that our eyeballs see, consider, and ultimately buy their products. Advertising is nothing new, but the omnipresence and sophistication of marketing continues to increase. With the popularization of AI applications, the way we shop is once again changing.
When Google reorganizes their search results page and adds AI-driven summaries, the information we use to make purchase decisions will change. When Amazon uses AI-driven algorithms to order their search results, the things we buy will change. Like in The Devil Wears Prada, decisions made far removed from the everyday consumer trickle down and affect us profoundly, often without our knowledge.
This is Part I of a series related to the evolution of product search and discovery, focused on how AI is changing how we find what to buy - and how brands market to us. How do we decide what to buy?
My favorite way to model how consumers purchase is McKinsey’s consumer decision journey.
My main focus in this piece is on the Active Evaluation stage. Imagine you are buying a dress for a friend’s wedding. At this stage, you may have a few brands or retailers in mind. During active evaluation, you’ll eliminate some options (too expensive, takes too long to ship), but you’ll also add some new options as you browse, talk to friends, or see other marketing stimuli. There’s a lot of money to be made at this stage for brands and for marketing platforms that court brands. As a result, during this stage, you are inundated with marketing. You may feel like the dress you looked at two weeks ago is following you around the internet - popping up on Instagram when you swipe through friends’ stories, then showing up in the banner of a news article. This is the window of opportunity for brands to convert your purchase, and they don’t want to miss it!
Active evaluation: How we search today
When searching for a dress for a friend’s wedding, what is your first stop? According to a recent survey of over 30k people worldwide, roughly 1 in 3 start on a marketplace, like Amazon. Retailer and brand websites and search engines are also popular starting points.
What does this have to do with AI?
Everywhere you shop (online, that is) is in an arms race to improve their search experience so you start your shopping journey there and keep your dollars there. If you are Amazon, you are very happy that 1 in 3 customers who needs a widget goes straight to Amazon and types “widget” in the search bar. They aren’t even looking at Walmart or Home Depot! What a win. However, if Google Shopping gets really good at finding me widget options, I might start exploring those options and Amazon may lose those sales.
Google: Recent product discovery innovation
In Part I, I’m focusing on some of the ways Google has implemented AI to change its search and discovery-related offerings in the past year. In Part 2, I’ll focus on other companies and features.
Google Shopping
Ahead of this holiday season, Google’s Shopping platform got several AI-driven upgrades. They added an AI overview summary provides explanatory text that is meant to help the user find the right product. They’ve also added more tailored product recommendations that are thematically organized. Between the overview text, sponsored results, and other recommended results, you need to scroll quite far to see organic results - posing a challenge for brands.
Immediately I see several issues with this experience.
Nonsense filters and categories. An “advantage” of the AI-driven experience is that the filters and categories are dynamic to my search query. The goal is to make it easier to find the right product by weeding out irrelevant search parameters. But this output doesn’t do a good job. For example, the result gives me the option to view mini dresses, which isn’t relevant to my search - “floral dress for formal wedding in May.” Formal dress codes imply a longer length. It gives “wedding dresses” as a category option, which is irrelevant as I’m clearly searching for a wedding guest dress. This adds up to lots of wasted real estate on the page and wasted opportunity to personalize the experience.
Text is not useful. In this case, I am looking to transact for a specific occasion, and I have specifically asked for formal, floral dresses. The text blurb is randomly discussing fabrics and vague dress code information. While this may be helpful in other contexts, in a commerce-oriented situation it is a distraction. There is a fine line between helping consumers make decisions and overloading with information - these text summaries haven’t yet found the balance.
Lack of personalization. The biggest missed opportunity here is the lack of personalization in these results. I am signed onto my Google account while making this search. Instead of showing me recommended options for “maxi” and “midi” length dresses, Google could show me sections like “Your Favorite Brands” populated with brands I’ve shopped or browsed before. They could prioritize filters on the left with retailers that I frequent. They could tailor the results to price points that I typically shop.
Shifting consumer behavior
The filters at the upper left hand corner - Get it fast, On Sale, Used, and Small business - remain constant for most searches. I love the inclusion of the “Used” filter as it makes it easy to view relevant secondhand items alongside new options. Similarly, the ability to seek out small businesses to shop from is a worthy objective. Yet when I look for “lipstick” with the small business filter checked, I see results from Burt’s Bees, L’Occitane, and Charlotte Tilbury. Are these really small businesses? This illustrates how a small product decision from Google has the potential to shift a lot of dollars to certain brands, likely at the expense of others. I’m curious to learn how consumers interact with these filters, and whether they become more dynamic as the AI experience matures.
The current experience is a sneak peek at a future where you can conduct your entire “active evaluation” without leaving the Google Shopping results page. This could be a good thing for Google, for the brands that do a good job ranking well, and (hopefully) for consumers who will save time and money.
As the owner of a huge amount of consumer data, including purchase and browsing history, Google seems well-positioned to provide a truly personalized shopping experience. I don’t think Google Shopping is there yet, but I’m excited to see where it goes. Google Search Results Page - Addition of AI Overview
Earlier this year, Google added a section to their SERP that summarizes the results in a blurb format, powered by AI. The exact types of output you’ll see depends on your search query. But the upshot is these newer sections on the SERP deprioritize organic search results and may make it harder for brands to get visibility to their websites in the ways they have traditionally.
Google has already started to monetize some components in the AI Overview section, placing ads below the summary text snippet in some cases.
My thoughts on the new SERP
For the consumer, these AI overviews can be useful because you can avoid clicking into various websites, save time, and theoretically see the best information across multiple sources with a quick scan. I find myself loving this for quick, everyday questions, like “what are signs your baby is teething” (haha).
When it comes to commerce, the utility of this module is not so straightforward. AI overviews don’t show up on all explicitly purchase-oriented queries. For example, googling “women’s black loafers” does not yield an AI Overview. But questions like the above - “how to get wrinkles out of a jacket” can be purchase-oriented. Perhaps you decide you need to buy an iron or a steamer.
The UI of the AI Overview experience reads very “trustworthy” - the text is on a white background without obvious logos to indicate that it may be from a source that is trying to sell you something. As consumers, we have been trained to view AI output as an unbiased representation of available information. Even with ads labeled clearly, it’s easy to imagine that many consumers see inclusion in the AI module as a “stamp of approval.” As a brand, I would jump at the chance to test ads in this module as I’d expect them to convert quite well. I’d also try to figure out how to “hack” the AI Overview by refreshing my editorial content for relevance. It appears the SEO hacker community is already hard at work trying to figure this out.
Fun related resource: This blog is a 10 year look at how Google has evolved its SERP - it’s striking to see the changes.
What’s next
Parts 2+ of this series will look at other AI-related search experiences, including visual search, Amazon, Perplexity and other AI chatbots geared towards shopping, site search advancements, and more.
Things I enjoyed this week
Puck’s podcast Fashion People is a great source of inspiration for this newsletter. I have followed host Lauren Sherman’s reporting for many years, and it’s so fun to follow
in her new role at Puck!I can’t get enough of 2025 predictions. Here is one of my favorites, from
- it’s a great breakdown of storylines that will dominate in consumer this year, and what it means for brands
ICYMI
I wrote about everyone’s favorite retailer - Costco!
Thanks so much! And I’m obsessed with AI and how to use to improve consumer experiences.
This was such an amazing piece! My main ethical dilemma with AI and the shopping experience is how it strips away individuality. Companies like Amazon, for example, design their marketplaces to keep customers engaged for as long as possible. Their algorithms often prioritize expensive options or subtly nudge consumers toward spending more, making it difficult to feel like our decisions are truly our own.
As you pointed out, decisions made far removed from consumers trickle down to affect us profoundly, often without our awareness. With AI shaping not only what we see but how we evaluate it, the balance between personalization and manipulation feels increasingly blurred. How can companies ensure they respect individuality while maximizing profits?