TMAI #471: 🤖 AI Age: Bye SEO, Hello AEO! -P4

[The Marketing Analytics Intersect, by Avinash Kaushik] [1]

TMAI #471: BYE SEO, HELLO AEO! -P4

[The Future of "Search": Agentic AI.]

[ Web Version [2] ]

PAST: Type a few keywords and having the “Search Engine” guess

what you want by returning 4 Paid and 10 Organic “blue links.” You

click on many links to research, evaluate, follow rabbit holes on

Reddit, remember what you wanted, decide, buy.

PRESENT: You type in a paragraph, or five, to explain the layers and

nuance of your need, the “Answer Engine” does all the research,

clicking back and forth, and gives you the final answer, or very

close. End result: Saves you 70% - 90% of time & stress.

That’s the core difference between current default Google and

ChatGPT, Mistral, others.

A relevant example for today’s story on Agentic AI: It is

university time in the US. Tens of thousands of kids will furnish

their dorm rooms. They can do the Past way and spend an insane amount

of time searching and researching and choosing.

They can also ask Rufus:

_I’m moving into an NYU dorm room. Can you find me everything I need

to furnish it for a teen boy?_

It comes back in a sec, asks you for your BUDGET, ROOM SIZE,

COLOR/STYLE preference, and PRIORITY. You reply with: _$500, shared

dorm room, priorities are comfort and study, minimalistic style_. Two

seconds later… BOOM!

[Amazon Rufus Agentic Experience.]

[Right click, open image in a new tab for a higher resolution image.]

Clearly, I love my child too much to offer a $500 budget. 😊 Rufus

did it for $330! I just hit add to cart to what I want. Time saved,

stress reduced, problem solved in under five minutes.

Rufus, launched in 2024 [3], is Amazon’s work in progress Agentic

AI, it is the great Answer Engine for shopping.

A multi-purpose Answer Engine can also tackle this task. Here’s

Grok…

Really helpful answers. Recommended more items, for just $245. And,

fascinating, it categorized items I could split with the

roommate. 😊

Due to its expansive scope, Grok [4] did one better than Rufus: It

found that NYU has a no-nails or tape policy and said I should use

Command hooks. Also, pre-identified that most dorms lack AC, and asked

me to prioritize a fan. Etc.

Grok researched 186 web pages to build my custom table answer. Default

Google brings me 186 pages, ranked, and expects me to do all that

work. [Click here [5] to try my query in Google. What do you see? Try

it on GEMINI, Google’s Answer Engine. Is Gemini as helpful as

Grok?]

The user experience of Answer Engines, glimpses of Agentic AI like

Rufus above, is _fanfreakingtastic _for humans. They are shifting

from typing “keywords” to what I call _resolution questions_.

For businesses though: REDUCED VISIBILITY. FEWER RESEARCH VISITS.

LESS BUSINESS.

Urgently activating Answer Engine Optimization (AEO) ensures we can

recover some of the reduced visibility, lost visits, and reduced

revenue.

In TMAI #468 [6], I’d shared the four problems Businesses need to

focus on:

1. PRE-SEARCH. 2. SEARCH EXPERIENCE. 3. PAN-SEARCH. 4. NEW

ANALYTICS.

Premium #469 [7], shared 11 actions you should consider across

Pre-Search (Research, Owned Experiences), and Search Experience

(Winning Paid Answer Advertising).

Premium #470 [8] was entirely dedicated to the Most Important THING:

Content. Please print, read every week, act every day, the ten

recommendations in Search Experience (Winning Organic Relevance).

Nothing is more important to your AEO strategy.

Today, something futuristic. You should start thinking about it now,

start working on it in six months, ahead of the good Agents coming

online over the next 12 to 18 months.

3. FUTURE OF SEARCH: AGENTIC.

We are beginning to get glimpses that “searching” is likely to

lose humans entirely.

Ex: Rufus did almost of my work today. Soon, it will become a

full-fledged Agent by closing the _last mile gap_ by placing the

order for items, and get them delivered to the NYU dorm.

If Answer Engines shrink the work demanded of us by Search Engines by

a factor of 10, AI Agents who work on your behalf will shrink it by a

factor of 100!

To get a glimpse of this, watch this launch video [9] of ChatGPT

Agent from a couple weeks ago. The examples of Minnia and Sarah

wedding planning and laptop sticker creation are illustrative of how

many, many humans will “search” in the next 12 to 18 months.

Basically: No human. All machine.

A new phrase needs a definition, here's mine for Agentic AI:

IT PLANS, REASONS, AND ACTS TOWARDS A GOAL ACROSS MULTIPLE STEPS &

TOOLS.

Applied to commerce, that is:

*

Searching.

*

Evaluating relevance, quality and fit.

*

Comparing prices, inventory, and shipping options.

*

Concluding the transaction, and track shipment until delivery.

Answer Engines already do A, B, C, as long as we are good at asking.

[TIP: There is no such thing as Prompt Engineering. Just speak like a

human, type out all the assumptions that are in your head, be specific

when you can, treat the LLM as if it is its first day on the job.

Instruct it to ask you questions, for clarifications, if it needs

them. Be prepared to have your mind blown.]

D represents the last mile gap, representing: 1. Technical

feasibility, and 2. A psychological worry if we can trust AI Agents

with our credit cards (as with the internet, we will get over this

soon enough).

Exciting Agentic AIs are coming online. Kruti can book [10] cab

rides (and track them arriving), order meals, and handle insurance

payments – all end-to-end, in 10 Indian languages!

AI Agents, applied to marketing, pose a massive disintermediation

risk.

*

What happens to Affiliate Marketing?

*

Email Marketing?

*

What happens to Loyalty?

*

What happens to our “Paid Search” ads, are they now for robots?

*

If we do not have a strong Brand, are we dead as cheapest price wins?

(Yes. Brand Marketing has never been more important than now!)

So… Is your business ready for a seismic change _PAN SEARCH_? Is it

ready for Agents to replace Humans when it comes to doing business

with you?

If you are normal, the answer is NO! That is ok. The time to start is

now.

Items 1, 2, 3 your IT team should have started work on in Jan 2025.

Items 4, 5, 6 your Marketing team should start working on now.

All six items below fall into a new acronym: AAO – AGENTIC AI

OPTIMIZATION.

Businesses getting ready to do business with machines, in addition to,

but soon instead of, humans.

Here’s the picture in my head re acronyms…

[SEO AEO AAO]

Let’s AAO…

1. SHORT-TERM INTEGRATIONS.

Early experiments with Agentic systems involve moving the checkout

experience into Answer Engines.

Here’s a little bit more about plans by OpenAI in the Financial

Times [11], or, if you can’t access that link, a bit more

in ModernRetail. [12]

This is essentially like integrating your checkout into Google Search,

so people can just click buy on Google. The short-term decision:

_Would you like to do the same for the major Answer Engines?_

Escalate this as a point of discussion with your IT team. The yes or

no decision should be made by your CMO. Then, if yes, decide when you

want to activate it.

In a technical level, short-term, you should focus on tactics like:

*

_Structured Data Fast Pass_: Add/update JSON-LD schema to top 20-30

revenue-driving PDPs and your main brand page.

*

_Entity Cleanup_: Ensure your brand and product names are consistent

across your site, Wikipedia, 3P retailer listings – AI Agents make

heavy use of these entity graphs.

*

_Setting up a Sandbox Agent Endpoint_: Build a simple endpoint with

clean product data so you can plug into a demo agent and practice.

Open a dedicated thread with your Legal & Compliance team to

understand how current policies, guardrails, consent, data retention,

need to be updated to apply to Agents vs. humans.

As Agentic AI evolves rapidly, more short-term possibilities will

arise. It is critical to be agile about this. It is extremely helpful

to have at least one human, or part of one human, dedicated to Agentic

commerce developments to help you be aware, evaluate them, and help

decide yes/no.

Now, let’s get to the really big stuff.

2. DIGITAL INFRASTRUCTURE & DATA FOUNDATION.

AI Agents rely on real-time identity resolution and context-rich

profiles. Oh, and, autonomous agents will transact on behalf of

humans, they’ll need tokenized payments, secure allowances and

secure transactions.

Agents abandon slow or error-prone endpoints immediately. If you

don’t have them already, monitor APO SLOs, like you currently

monitor checkout uptime.

If you have a monolithic commerce architecture, the front-end visual

layer and backend business logic are tightly coupled, this could pose

a major barrier for Agentic interactions.

HEADLESS COMMERCE allows AI Agents to bypass the human-facing layer

and interact with business logic through APIs. This frees up Agents to

execute tasks and adapt experience in real time without the constraint

of a rigid presentation layer.

COMPOSABLE COMMERCE takes this a step further by breaking down the

backend itself into a collection of individual, independent, API-first

microservice. Ex: Search, shopping cart, promos, payments all exist as

separate services that communicate via APIs. This massively modular

architecture is ideal for Agentic systems.

Does your awesome ecommerce tech stack support these new needs?

3. PREPARE YOUR SITES AND MOBILE APPS FOR AGENTIC INTERACTIONS.

Three important areas of focus...

CONTEXTUAL METADATA: Machine-readable, structured product data will be

even more critical. Rich metadata, attributes like size, material are

common, that’s great, but remember how LLMs think, you will need to

ensure lifestyle context (“for mountain biking”) will become

critical. Help agents understand product relationships, bundling

opportunities, and customer segment relevance.

SEMANTIC PRODUCT INFORMATION:  Detailed, machine-readable, product

descriptions that include technical specifications, use cases,

compatibility information, and comprehensive features in standardized

formats.

PROGRESSIVE ENHANCEMENT: Ensure your websites function effectively

for both visual browsers (humans) and programmatic access, so agents

can complete tasks regardless of JavaScript availability.

Print this and post it on every wall of the IT department: APIS AND

INTEROPERABILITY!

For Agent AI sent to transact effectively, your platforms must offer

robust APIs for 3P agents.

[NOTE: Work you have done to activate Google Merchant Center,

Shopify’s Storefront GraphQL API, will give a nice foundation to

jump forward.]

4. AGENTIC ORGANIC SEARCH.

Ensure Content (and AEO) is for Agents, not just humans.

We are influenced by marketing tricks like visuals, emotional copy

etc., these matter a lot less for an AI Agent – it needs clean,

semantic, AI-optimized copy.

Your IT team should already be testing how AI Agents currently

interpret sponsored tags, pricing, endorsements, next best offer, and

optimize to agent-preferred layouts – ex: crisp, canonical specs and

comparison tables, best for statements for use cases, compatibility

matrices, _yada, yada, yada._

Technical AEO will also need more focus: Structured snippet

optimization, voice search compatibility, local and contextual

signals, that help agents make optimal recommendations re your

business.

Bits that we currently hide like Returns policy will need to get

exposed because returns is a huge concern of humans, and hence their

Agents will abandon you if they can’t find your returns policy

clearly spelled out on every single page of every single

product/service. Or, if they can’t translate easily that “Final

Sale” for a “Last Chance Item” means no returns.

5. AGENTIC PAID SEARCH.

It is not entirely clear at the moment, but I anticipate that

today’s Paid Search will become API-driven as well. Ads will target

Agentic interfaces (not just humans). You will have to develop direct

incentives and visibility with AI Shopping platforms (like Rufus

above, Shopping Hub, Sparky, Kruti etc.).

Ads are likely to move inside (Google has previewed this). You might

win or lose before an answers page is visible. Invest in feed quality

(titles, attributes), audience signals, and creative that maps to

_resolution questions_ and considerations.

Microsoft’s preview of ads in Copilot indicate that your paid ad

assets need to be able to power dynamic comparisons (specs, UGC, FAQs)

vs. static banner creatives (or creatives built by your creative

agency). Activate this.

[Premium Member Bonus: TMAI #436: Creative Strategy: Performance vs.

Brand Ads. [13]]

These are unknown at the moment re “Paid Search”:

*

Will we be bidding to Agent attention?

*

What will Agent-centric ad formats look like (functional, not

sexy?)?

*

Beyond Search, how will we create advertising inventory that Agents

can access programmatically during their _plan, research,

act_ process?

*

What “Agent trust signals” will we have to create to build

credibility so that Agentic systems will recognize and value us

higher? (Quality Score [14] does this at a tactical level today.)

*

Scary one: Will we be undertaking _Brand Marketing _exclusively

focused on algorithms and Agents to influence their decision making re

placement, value, priorities of our _Performance _ads? (Yes, only a

matter of time. We just don’t know how yet.)

My objective in sharing these with you is to move some of your

_unknown unknowns _to _known unknowns_. Then, as an organization that

cares about its survival, you’ll know what you don’t know.

Hopefully, as recommended above, you’ll also hire someone to be

explicitly responsible for evolution of AEs and Agentic AIs, who will

identify actions as _known unknowns _become _known knowns_.

6. SERVER-SIDE MEASUREMENT.

Our Analytics tools (Google, Adobe) will evolve to accommodate AI

Agents. In the version of the tools we use today, AI Agents look just

like the bots that our measurement platforms have been optimized to

hunt down and destroy!

My speculation is that there is a real possibility that as more

conversions arrive via APIs, we might have to move to server-side

measurement to get clean signals.

Just be aware of this, we can cross this bridge when we get to it.

For the humans visiting our websites via Answer Engines, you should

already see partial data for them in your GA 4 reporting. Default

Channel Group = Referral. Source = chatgpt.com. Medium = Referral.

For others the source

=  gemini.google.com, perplexity.ai, claude.ai etc. In some

instances, ChatGPT Search as an example, you’ll see Source = Direct.

We will likely use a combination of server-side measurement (bots) and

normal web analytics (humans) to stitch the complete story together.

[NOTE: We’ll do a deep dive on new analytics next week.

Methodologies & KPIs.]

HOW PREPARED IS YOUR IT TEAM?

My sense is, it would be helpful for your CTO/CIO to have a simple

guide to assess where your company is re AAO readiness.

I’ve taken the IT specific actions, and classified them into Legacy

(Level 1), In Progress (Level 2), and Agent-Ready (Level 3).

[AAO Readiness]

[Please click here [15] for a higher resolution version. If you prefer

the Excel table, just reply.]

What’s your company’s Readiness Level?

NEXT WEEK.

Analytics!

There is no Webmaster Tools. There are no referring “keywords.”

There are currently no incentives in place for Answer Engines to share

data with us (they don’t run ad-based revenue models – so far).

To fill this gap, we’ve seen a number of _competitive

intelligence _tools spring up. They also have no direct access to the

data from the Answer Engines, but they have figured out different

approaches to _hack_ the LLMs to get some data.

Let’s separate the wheat from the chaff next week, so that we can

more intelligently inform our content strategies and organic

relevance.

BOTTOM LINE.

Agentic AI is almost there, it will take a year or so to go mainstream

for the _people on the edge. _So, you have some time to put in place

an effective AAO strategy and execution plan.

I want to emphasize this point: Even if all of Humanity does not use

Answer Engines tomorrow, anything you do for AEO WILL BENEFIT your

SEO immeasurably. Even if all Humanity does not switch to using

Agents, AAO WILL BENEFIT your organization’s tech stack, agility,

and readiness for whatever is ahead.

Don’t accept excuses.

Carpe diem.

-Avinash.

PS: This evening I was sharing with my boss, Tapestry CFO Scott Roe

[16], [16] my assessment of Agentic AI taking over from Humans, and

he said: _Oh, that’s just like The Forbin Project, from

1970._ Here’s the movie’s trailer [17]. WDYT? 😊

Thank you for being a TMAI Premium subscriber - and helping raise

money for charity.

Your Premium subscription covers one person. It's fine to forward

occasionally. Please do not forward or pipe it into Slack for the

whole company. We have group plans, just email me.

[Subscribe [18]]  |  [Web Version [2]]  |  [Unsubscribe [19]]

[20]

[21]

[22]

©2022 ZQ Insights  |  PO Box 10193, San Jose, CA 95157, USA

Links:

------

[1] link

[2] link

[3] link

[4] link

[5] link

[6] link

[7] link

[8] link

[9] link

[10] link

[11] link

[12] link

[13] link

[14] link

[15] link

[16] link

[17] link

[18] link

[19] link

[20] link

[21] link

[22] link