From SEO Maturity to AI Discoverability: Heather Physioc on What Needs to Change

Search is changing what discovery means for retail brands.

I have known Heather Physioc for a long time, and I still remember her original work on SEO maturity. It gave a lot of us a better way to talk about what "good" looked like inside a large organization: not just better title tags, cleaner technical foundations, or more content, but the people, process, measurement, and buy-in needed to make search work properly at scale.

That was the starting point for this conversation. Heather is now Chief Discoverability Officer at VML, and her role reflects where the industry has gone. Search is no longer neatly contained inside classic rankings and blue links. People discover, compare, research, and buy across AI answers, retail search, marketplaces, social, reviews, paid search, local, and product feeds. The old SEO maturity model is not dead, but AI discovery is forcing us to update it.

˙✧˖ AI-written summary

Below is an AI-assisted summary of the webinar conversation. This is not a word-for-word transcript but is included to help you find the key parts of the conversation.

Search maturity needs a reset

Heather's original SEO maturity model was built around the idea that search success depends on more than technical fixes. Mature teams need strong foundations, good content, measurement, organizational support, and a way to keep improving.

That is still true.

The change is that AI discovery makes some of those areas much more important, and exposes weaknesses that old SEO dashboards could hide.

In the old model, a team might look mature if it had clean crawlability, decent rankings, some content governance, and enough internal influence to get fixes shipped. Now, that same team may still be underprepared if its product data is messy, its brand messaging differs across channels, its reviews and third-party descriptions contradict its own site, or its reporting still assumes every journey ends in a clean last-click conversion.

Heather described the buying journey as a much messier environment now. People may ask an AI assistant for options, read a summary, check reviews, compare products in a retail surface, watch social content, and only later arrive at the site. Some of those touchpoints look like search. Some do not. Some generate measurable traffic. Some do not.

That is why search maturity needs a reset. The foundations remain, but the scorecard has changed.

Technical SEO is becoming strategic again

There was a period when people talked as if technical SEO was becoming less important. The CMS would solve it. Google would render everything. Platforms would handle the basics.

That was always too optimistic.

Heather's point was that the basics are now assumed. Crawlability, indexability, site speed, mobile friendliness, and clean rendering still matter, but they are the floor. The next layer is machine readability.

That includes:

• structured data
• product feeds
• taxonomy
• content architecture
• entity relationships
• product attributes
• category structures
• canonical sources of brand and product information

This is where technical SEO becomes much more strategic.

AI systems need to understand what a brand sells, what products are available, how they differ, who they are for, what reviews say, what attributes matter, and whether all of that information is consistent across the web.

That is not just "add schema". It is a deeper question: can machines understand your business accurately?

For large retail and ecommerce sites, this is a serious operating challenge. Product data may sit in one system. Content in another. Reviews somewhere else. Store data elsewhere. Paid teams may use one feed structure, organic teams another, and marketplaces may have their own version of the truth.

Humans can sometimes cope with that mess. Machines may turn it into confusion.

This is also where structured data and experimentation meet. SearchPilot's SEO testing platform supports controlled tests on titles, metadata, on-page content, structured data, and page templates, so teams can find out which machine-readable signals actually change performance.

The brand data core

One of Heather's strongest recommendations was that enterprise teams need to build a brand data core.

That phrase is useful because it takes the conversation beyond SEO tags and dashboards. A brand data core is the shared source of truth that tells machines and people who the brand is, what it offers, what it should be associated with, and why it should be trusted.

This matters because AI discovery does not pull only from your website.

It can synthesize information from owned content, earned media, social platforms, reviews, marketplaces, local listings, retail partners, product feeds, and third-party databases. If those sources disagree, the brand becomes harder to understand.

A simple example: your website says a product is sustainable, your product feed does not include sustainability attributes, reviews talk mostly about price, PR coverage focuses on fashion, and retail partners use old product descriptions. Which version should an AI system believe?

The answer is not obvious.

That is why this becomes an organizational problem, not just an SEO problem. The brand, product, ecommerce, social, PR, paid, SEO, and data teams all contribute to the picture. The more fragmented that picture is, the harder discoverability becomes.

The goal is not to centralize every decision. That is unrealistic in most enterprise companies, and probably not desirable anyway. The goal is to create enough shared structure that every channel reinforces the same facts.

Product feeds are not just plumbing

For retailers, one of the biggest mindset shifts is around product feeds.

A lot of organizations still treat feeds as ad ops plumbing. Necessary, technical, and slightly boring. Something that keeps Shopping campaigns running.

Heather argued that this view is out of date.

Product feeds now sit at the centre of paid shopping, organic shopping, Merchant Center, product search, and AI-assisted recommendations. They are not only a way to pass products into an ad platform. They are a machine-readable description of what you sell.

That matters a lot as discovery gets more agentic.

Imagine a shopper asking an AI assistant to find the best washing machine for a small apartment, with quiet spin, quick delivery, installation, and recycling of the old appliance. The system needs structured information to compare options. It needs dimensions, availability, delivery promises, reviews, energy ratings, service options, and product attributes.

If your data is thin, inconsistent, or spread across disconnected systems, you are making that comparison harder.

This is why product feeds, taxonomy, structured product data, reviews, availability, and pricing signals are moving from the back office into the search strategy conversation.

For ecommerce teams, this is also a good place to connect to SearchPilot's work on Merchant Center Testing and controlled experiments across product and category templates.

Visibility metrics are not enough

Heather was careful about the new wave of AI visibility metrics.

It is useful to know whether a brand is mentioned, cited, or surfaced in AI answers. That kind of tracking can help teams see whether they are present in new discovery surfaces.

But it does not answer the harder questions.

What changed because of that visibility?

Did it influence demand?

Did it improve trust?

Did it move revenue?

Did it create better qualified visits?

Did it change what people searched for next?

Did it reveal a data or content gap the team can actually fix?

A dashboard that says "you appeared in 43% of prompts" may be interesting, but it is not automatically useful. The mature team asks what to do with that information.

Heather was also cautious about sentiment metrics. The issue is not that sentiment is irrelevant. It is that sentiment analysis can be blunt, especially when it tries to compress complex language into positive, neutral, or negative. The better question may be: what associations are being made around the brand, where do those associations come from, and which ones can we influence?

That is more actionable than treating AI visibility as a new rank-tracking report.

For a deeper dive on ecommerce KPI drift, especially the pattern of traffic falling while revenue rises, see our piece on whether SEO teams are chasing outdated KPIs.

Last-click reporting is breaking

Last-click attribution was never perfect, but it is becoming less complete.

AI answers, zero-click searches, personalized results, agentic shopping, and fragmented discovery journeys all create moments of influence that may not appear as a clean organic visit.

A shopper might see an AI-generated comparison, ask a follow-up, check Reddit, watch a video, see a product listing, and later buy direct. Under a traditional model, much of that influence disappears.

We have written separately about the same problem as When is a click not a click?, because fewer clicks can still mean more valuable clicks if more research happens before the visit.

Heather's point was not that teams should throw away every existing metric. Rankings, traffic, clicks, revenue, and conversions still matter. For retailers, there is still a transaction somewhere. Someone still buys the sofa, the trainers, the fridge, or the holiday.

But mature teams should stop pretending the old scorecard captures the whole journey.

One pattern we discussed is that organic traffic can fall while organic revenue rises. That is not always good, and it needs careful interpretation, but it shows why traffic alone is not enough. Some lower-value informational clicks may disappear into AI summaries, while the remaining visits become more qualified.

The right question is not just "did traffic go up?"

The better question is: what changed in the journey, and what happened to the business outcome?

Apply caution to synthetic promptsThe same caution applies to synthetic prompt tracking.

Heather is skeptical of dashboards that run a set of prompts and imply they represent the market. LLMs are not deterministic. You can run the same prompt many times and get different answers. Different users have different histories, locations, contexts, and follow-up questions. We do not yet know how real people are using these systems at a granular level.

That is why Will has been using the phrase "search volume one."

The query may look similar, but the actual experience is personal. The follow-up question makes it even more personal. Measurement has to adapt to that reality.

The technical model matters here too: LLMs do not rank pages in the old SEO sense, so AI visibility reports need to be interpreted with care.

For practical tracking guidance, the related SearchPilot article How does AI traffic show up in analytics? is a useful companion piece for teams trying to understand what AI referrals look like in analytics and logs.

Silos are now a discoverability problem

People have been talking about silos for years. Usually it sounded like a nice organizational improvement project.

AI discovery makes it more urgent.

SEO, PR, social, brand, product, retail media, local, marketplaces, reviews, engineering, and data all shape what machines understand about a brand. If those teams are disconnected, the signals become disconnected too.

That matters because AI systems do not respect org charts.

They do not know that the PR team owns one message, the ecommerce team owns another, and the feed team uses a third naming convention because the platform requires it. They just see a messy collection of signals and try to synthesize something from it.

Heather said a lot of her work now is helping enterprise clients bring those groups together and decide where to start. That feels right. AI has captured leadership attention in a way SEO sometimes struggled to. That creates an opening for search teams, but it also raises the standard.

A mature answer to AI discoverability is not "we need to write some GEO content."

It is: we need a shared operating model for how the brand is represented, measured, tested, and improved across the surfaces where discovery happens.

What mature teams will do differently

The top 10% of teams will not treat AI readiness as a side project.

They will not create an "AI visibility" dashboard, run a few prompts, and call the job done.

They will do the less glamorous work:

• clean up product data
• align brand messages across channels
• improve structured data
• connect feeds, PDPs, PLPs, reviews, and content
• bring SEO, PR, social, product, paid, data, and engineering into the same conversation
• test faster
• report against business outcomes
• build feedback loops from what they learn

This is not a quick checklist. It is a maturity shift.

The most mature search teams will not be the ones with the longest list of AI tactics. They will be the ones with the cleanest systems for learning.

For a broader discussion of turning enterprise SEO data into action, the Patrick Hathaway conversation on what teams need beyond crawl data is a useful companion piece.

Agentic shopping is coming, but the shape is still unclear

Heather and I also talked about where shopping might go next.

Universal carts and agentic buying are still early. Some versions feel like toys right now. But the direction is not hard to imagine. Platforms want to keep users in their environments. AI assistants want to complete tasks, not just answer questions.

For retailers, that creates big questions.

Where does the transaction happen?

Who owns the customer relationship?

How does attribution work?

What happens to the website if an assistant can compare options, choose a retailer, and complete checkout on behalf of the customer?

We do not know the final shape yet. It might be platform-led checkout. It might be API-based commerce. It might be a general agent acting like a user, visiting a retailer's site and buying through the normal checkout flow. It may be all of those things at once.

The practical takeaway is the same: retailers need their product, pricing, availability, delivery, reviews, and brand information to be machine-readable and consistent.

The cleaner the data, the easier it is for any future buying agent to understand and recommend the product.

What search leaders should do now

For search leaders, Heather's advice was practical: do not try to solve everything at once.

Start with a priority area. Pick a category, product line, service proposition, or customer need that matters commercially. Then look across the ecosystem and ask:

• Does our product data support this?
• Does our content support this?
• Do reviews reinforce it?
• Do third-party descriptions match it?
• Does structured data make it clear?
• Do social and PR signals support it?
• Do our internal teams agree on the story?
• Do our metrics tell us whether this is working?

That is a better starting point than "optimize for AI."

It is also easier to fund. Leaders can understand a specific commercial priority. They are less likely to fund a vague AI readiness project that never ties back to outcomes.

Teams that are ready to test this can also explore SearchPilot's GEO A/B Testing, where the focus is not abstract AI visibility, but whether AI-generated answers and overviews lead buyers back to the site.

Put Search in Control Mode with SearchPilot

This conversation with Heather came back to a familiar problem: search is often one of the biggest channels in the business, but one of the least understood.

AI discovery makes that uncertainty worse. The surfaces are changing. Measurement is getting messier. Buying journeys are less linear. Product data, feeds, content, structured data, and brand signals all matter more.

The answer is not to guess harder.

SearchPilot helps enterprise teams make SEO and GEO testable. We run controlled experiments across category pages, product detail pages, templates, navigation, internal linking, content, structured data, Merchant Center surfaces, and AI-influenced journeys, then help teams understand what moved and what the result means.

For retailers, that means testing the systems that increasingly shape discoverability: product data, PDPs, PLPs, feeds, content modules, entity signals, Merchant Center surfaces, and AI-influenced journeys.

The goal is not to predict every future interface. It is to build a way of working that can keep learning as the interfaces change.

For examples of this in practice, see the SearchPilot customer stories from M&S, adidas, Omio, Flight Centre, Skyscanner, and The Stepstone Group.

Search is your biggest channel and least understood. Take it out of react mode. Put it in control mode.