At the recent Google Marketing Live event, the search engine described how their data was showing that AI-powered search is pushing people faster into the most valuable state in marketing – confident, informed and close to buying.
Google has been talking about the “messy middle” in customer journeys for almost a decade. The is the non-linear space between trigger and purchase where people search, compare, hesitate, validate, and eventually decide and commit.
Now the tech giant is suggesting that the middle is not quite as messy for people that are using AI-powered search. Their data says that AI Overviews and AI Mode users make decisions faster. With the guidance and recommendations offered by the AI, meaning they feel more confident to pick and purchase.
If this is true, the implications are significant. As this will narrow the number of touchpoints and opportunities for persuasion on the consumers path to purchase. For brands, this means that showing up in AI engine answers (so called GEO or AEO) becomes all the more important.
Potentially condensing the journey further Google has shared their plan for a re-imagined ‘intelligent search box’ which can drop users into AI-powered experiences.
Interestingly they also talked about a generative user interface where search can build a custom tool or interface on the fly, rather than handing you a list of links. This again has significant implications for brands that use search to generate traffic and sales, and once again means that being the source the model trusts matters more when AI is synthesising answers and interfaces and tools are being generated on the fly. Visibility in an agentic search world is less about ranking on a list of links and more about being retrievable, structured, and credible.
As the web becomes more agentic, we are starting to see a range of different strategies being adopted when it comes to AI shopping agents.
JD Sports are an example of a retailer integrating agentic commerce into their site directly. Partnering with Stripe to deploy the fintech’s agentic commerce suite which means they can sell through AI chat alongside browsing and traditional search.
Shopify have taken this further, recently switching agentic storefronts on by default for over a million merchants, including brands like Glossier, SKIMS and Vuori. This means that agentic capabilities on Shopify will now be opt out rather than opt in.
Amazon is interesting, as it’s doing two things at once. On the one hand, it’s blocking the outside agents, pulling around 600 million products from ChatGPT’s results and they also sued Perplexity over its scraping its agentic shopping agent scraping the site.
On the other hand, it’s building and deploying its own agentic capability. They recently announced that they will be folding their Rufus chatbot into Alexa for Shopping, a new assistant that sits directly in the search bar that draws on what it knows about a user’s purchases and preferences to act as a personal shopper. Alexa Shopping can compare products, track price drops, automates reorders, and through a ‘Buy for Me’ feature can even buy from third-party sites on a user’s behalf. Amazon’s big bet is clearly to keep rival agents out and make its own agent the consumer’s default.
These three examples – brand adoption, platform adoption, and a walled garden approach show how rapidly but also broadly agentic commerce is developing and it’s likely that each of these strategies will play at least some part in the future of digital commerce.
Google’s I/O conference also saw a significant agentic commerce announcement with the upcoming launch of Universal Cart. A single cart that follows the customer across Search, Gemini, YouTube, even Gmail.
If a consumer adds something to the Universal Cart while browsing, it will go to work in the background, tracking price drops, flagging when stock returns, and comparing payment perks.
The Universal Commerce Protocol lets people check out within Google, or transfer to a merchant site to finalise the transaction. The opportunity for the consumer is reduced friction but there are clearly some significant potential gains here for Google in terms of shopper data.
Once again, it’s a trade-off for brands between reach and handing Google ownership of more of the customer relationship and data. It’s unlikely that platforms like Amazon and other retailers will be willing to sacrifice much in the way of customer data so it will be interesting to see whether Universal Cart able to gain significant traction.
Matthew Prince, the CEO of digital security and content delivery business Cloudflare, recently noted how we have already crossed a notable threshold with automated bot traffic surpassing human traffic online for the first time in internet history.
According to Cloudflare bot traffic now accounts for around 57% of online traffic (measured in HTTP requests). Prince had projected only recently in March this year at the SXSW conference that this crossover wouldn’t happen until 2027, but this has been dramatically accelerated he says, specifically by AI agents rather than traditional search engine indexers.
Part of the reason why agent traffic has increased so substantially is that when an agent responds to a request from a human, for example for planning a trip, it can scrape thousands of sites very rapidly which dramatically increases the number of page requests.
This raises interesting questions for content producers. Nicholas Thompson, CEO of The Atlantic, has noted that content owners and producers have a choice between embracing their content being scraped and used by AI agents in return for visibility, or blocking agents from crawling their site to protect their business model.
Interestingly The Atlantic has partnered with a company called Parallel AI (founded by former Twitter CEO Parag Agrawal) which provides a tool that tracks exactly how much their content is used by AI agents with the potential to earn revenue when it is.
Another interesting example of content owners responding to changes in the digital landscape is The Economist, which is preparing for what its head of generative AI calls a world with two versions of the web, or a two-track internet.
This calls for two sites. The one we already know, built for human reading, engaging editorial, visual design, and narrative. The other is stripped right back, restructured into clean questions and answers that an AI agent can parse without tripping over features like carousels and feature art.
The logic behind this experiment is that discovery will increasingly start with an agent acting on someone’s behalf rather than a search box or the homepage. Creating two versions of their site serves both a human and an AI agent audience in a way that makes The Economist discoverable by agents but also readable for humans.
For now, The Economist is testing this on marketing and sales pages that sit outside the paywall, but it raises a question that will likely come up much more – who exactly are we designing for? Humans or agents or potentially both.
Join us for: Social media beyond the ‘like’ with Lisa Hale, Head of Brand Communications at Specsavers // 30th July, 1pm BST