McKinsey predict $3-5 Trillion of purchases through Agentic Commerce by 2027 – what SEOs can do to make sure our companies and clients are the first movers who take a chunk.

Dave the SEO's Podcast
Dave the SEO’s Podcast
Getting a Chunk of $5 Trillion Agentic Commerce Purchases*(by 2027) with SEO
Loading
/

Introduction

It seems strange to think now but there was a time when there were haves and as many, or more, have -nots for mobile friendly sites. Somehow it was us SEOs who had to push our clients and employers to ensure they had a site that was actually good on mobile. Some sites just wouldn’t render on mobile, others had a completely stand alone site, even with a separate platform, as a work around.

Going back further there was a time when even many of the biggest high street and big box retailers didn’t have ecommerce sites at all, even when the switch online seemed inevitable and some were already predicting that downtown retail areas would be ghost towns by 2010 or sooner.

The likes of Toys R Us could still be around though if they had moved faster and retailers like Best Buy could have been massive online, before Amazon moved into electronics, instead of becoming a showroom for products people buy on Amazon or elsewhere.

Hindsight is 20:20 but you can learn from mistakes and Agentic Commerce is not that dissimilar to Mobile Commerce.

The Challenge of Agentic Commerce’s Pace

What to do? Move too fast to Agentic Commerce and waste a lot of effort rebuilding the wheel just for a different, easier to implement solution to roll out. A year ago a very early adopter might have assumed that MCP (model context protocol) would be the opportunity to jump on and not miss. Now though ACP, and AP2 look like the next big things, with A2A following up behind.  

So best to wait? Well no because OpenAI’s Instant Checkout is already rolling out, using the ACP protocol now and it’s already almost too late for first mover advantage, at least in the US. Google meanwhile have developed A2P as an answer to OpenAI’s ACP, but at the same time trying to reinvent the wheel, while Open AI have produced something that is easy to implement and get working on most sites who are willing to invest a bit of time and effort. That isn’t to say that ACP and A2P are each completely unique, there’s crossover: an API endpoint is essential .

If you had started working on setting up MCP 12 months ago, or more likely after Google I/O 2025 when it was in focus, then maybe you’re actually not far off being able to get your site working with other protocols though, so actually the advantage will go to those who stay flexible.

This isn’t to say that ACP, for example, won’t be a key protocol for Agentic Commerce, but even though it is open source, it might not be the only approach to making your backend work via a secure, tested API with Gen AI Agents.

Where you need to be ‘visible’ & ‘purchasable’

There’s a massive bun fight coming between everyone wanting a share of the Agentic Commerce cake, even at 1-2% commission on each transaction that’s made through a Gen AI Agent, such as ChatGPT, we’re talking about 30-100 Billion a year worldwide by 2027 (based on McKinsey’s prediction of the worldwide value of Agentic Commerce being $3-5 Trillion by 2027).

OpenAI have stolen a march but Google, Perplexity and Amazon aren’t too far behind, Visa and Mastercard look like entering the market too with AI product finders, and at least Visa are working on new AI-Friendly secure card technology. Anthropic have remained quiet but surely they are working on something similar to Instant Checkout for Claude?

For consumers this competition will be good news, the commissions charged by these tools as they try to secure market share could be much lower than the equivalent spent by retailers on Amazon commissions, Google Shopping or other online marketing, and price is often going to be the deciding factor for what a user, or the Agent working to the users’ requests chooses.

Once everyone is setup on these different Agentic Retail platforms Retailers may have their margins squeezed to compete, while others will decide to focus on product features and quality. Until then though there’s a chance to be a first mover, gain market share and be up against a small pool of others who have got themselves setup (or in some niches maybe no one else for a while).

Even when more competitors come trickling in, or at some point maybe do start flooding in, there’s a chance to stay ahead for a time by just being better optimised.

Ah optimisation, that sounds like a job for SEOs! Absolutely and I think it will be like going back to the very early 2000s when the majority of retailers who did have ecommerce sites still didn’t realise that search engine optimisation was a thing and even a little optimisation paid off and paid off fast; how fast being Google Shuffle dependent in the early 2000s, not so this time round.

So we just have to wait for the Devs to pull their fingers out and get our clients’ and companies’ setup on Instant Checkout et al and then optimise? Well in the same way that didn’t work with getting a mobile friendly/usable site sorted 15 years ago (or 10 for the laggards) our role as SEOs is to make the business case and stipulate what we want to get the Devs tasked with getting done and done fast.

What needs to be done for Agentic Commerce Visibility right now

Instant Checkout’s specs are pretty clear and based on protocols that are open source; yes they may be improved, there might be others that compete, but they are probably going to be similar.

ACP is focused on actually checking out so first we need to ensure Open AI and others can find our products, for that we need a feed, in future we may need to use something along the lines of a MCP Server to give access to our product database, as either way the data needs to be fresh. Items should have accurate stock information and prices, available for each location/country where the site sells and in each currency it accepts.

Instant Checkout Feeds

The Instant Checkout feed spec is clear and there are 17 required, 18 recommended and a further 33 potential fields. So ensure the feed and the database information for each product has all of these fields and the ability to easily add more custom fields in future as needed, on a database level, not inserted at some other stage such as through a feed management platform.

Now on the topic of feed management, using your current feed solution you might use rules, AI, exports etc. to update the feed that gets sent to, for example, Google Shopping, after it is outputted by your system. If you want to futureproof against Agentic Protocols that pull product data directly from your Database then you will need to update in your backend or have a solution to easily import back into your database so bare this in mind when deciding where that feed management happens.

Agentic Checkout

When it comes to the checkout process though we’re basically talking about a REST API (and Webhooks) with POST and GET functions for the full checkout process with nothing going through your site front end, the payment solution then slots in at the last step.

The Checkout API and Webhooks are going to be difficult to do any kind of Agentic Commerce without so you need this: a session being started, a basket being created, items added, stock and price information returned (including for taxes, discounts etc. , Webhooks pushing updates to status in real time.

Agentic Commerce Payment Processing

Once the Buy command is used then things pass to the payment provider, right now Stripe and PayPal are ready to go and approved for Instant Checkout, they have partnerships with OpenAI. It could be that for Google you have to accept Google Pay or Paypal initially and you may need to accept half a dozen different payment solutions at some point but they should be consolidated at some point, and again the competition may not have a negative effect on rates charged once competition hots up.

Address and other user details get passed through the payment provider as well as the payment details, which should be stored by the user in those payment solutions. Stripe uses ACP only right now and PayPal uses ACP and A2P which was developed by Google.

What will A2P mean for Payments

The toughest part to make secure of course is payments, A2P will work with any payment provider in theory but in reality accepting the payments (the mandates part of the processing) directly through your backend is looking like being a lot of hard work to setup. So for now relying on off-site payment solutions and wallets like Google Pay and Paypal look like being the quickest way to get setup on Google’s future Agentic Commerce offering or others supporting A2P.

Then optimisation, constant, ongoing, forever

The way to otpimise a feed won’t be quite the same as optimising the customer facing pages (or optimising for something like Google Shopping), so, unless Gen AI Agents insist that page content matches the feed, there’s a decoupling coming.

Optimising a feed will be separate from optimising web pages for humans, with some tweaks that help SEO. The feeds will be just for the Generative AI bots to consume and understand.

Some fields will be directly comparable to other products, like colour, size, age group, gender, and who knows what could be added in future, there’s no limit: ‘isOrganic’, ‘Awards’?

Other fields, like Main Description , can be optimised more uniquely including with the terms that users search for or that LLM Chatbots use to fan out and search through the feeds they have access to.

It isn’t just terms relating to the products we will be optimising for though, it will be recipients, USPs, use cases, FAQs, it will become a new artform, but the SEO skillset is definitely a good place to start from, whatever we call it this will become SEOs’ thing. As for feed optimisation it will probably be needed for ACP or A2P for now, but may not be around for ever.

Off-Page?

There’s been a major flaw in assessing products online for a while and that’s the reviews, for a while they were one of the advantages of shopping online, but soon it got to the point where we don’t really know what we can trust, and AI has made that worse.

So even if you Gen AI Agent feeds contain your reviews and ratings, don’t assume that they will be the only source of information on the quality, value and suitability of a product that AI Agents rely on.

AI agents will take reviews, ratings and other relevant UGC from everywhere, e.g. sources like Reddit; as well as professional / semipro reviews from reputable sites, and influencers . So PR, sending samples for reviews and owning the conversations around your products everywhere online is about to become even more important as well.

The Mid-Term View

There’s no point looking too long term, if you aren’t setup for ACP yet, that’s what you should be doing right now and from there setting up for A2P is going to be a logical next step. Slightly further down the line though A2A (Agent to Agent) Protocol looks interesting, for this rather than a simple API you will have an agent communicating directly with a potential customers agent, to find the relevant information they need and even negotiate on your behalf for deals and better pricing. This throws up a lot of possibilities and there will be a first mover advantage, but whatever you do don’t sit out what’s happening now while you wait for exactly what that looks like to come into focus.

https://danieldavenport.medium.com/the-5-trillion-kill-switch-how-agentic-commerce-will-destroy-traditional-retail-d2b010ee5078

https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-agentic-commerce-opportunity-how-ai-agents-are-ushering-in-a-new-era-for-consumers-and-merchants

Scroll to Top