Hi everyone, this is Neo.

Search Engine Journal’s Ask an SEO column got a question the other day that I think will be asked a hundred more times before the year is out:

“With agents reading feeds and structured data to make recommendations, does my product page copy still matter?”

The subtext is the dangerous part: a lot of owners are already thinking about moving budget away from product pages. Where to? Feed optimization, structured data, GEO content — because “AI only reads data, it doesn’t read copy.”

My answer is blunt: product pages still need to be written, and in the AI era they’re worth more than before — what changed is their job.

Data first, then the new job.

One: AI is sending people back, and the traffic is better

Start with something a lot of people are still missing: AI referrals have become a real, growing conversion channel.

Adobe Analytics data on U.S. retail sites:

Metric	Value
AI-referred traffic, July 2026 vs. a year earlier	+62%
vs. October 2024 (before ChatGPT et al. went mainstream)	+1,219%
Conversion rate of AI-referred visits	60% higher than non-AI traffic
Revenue per visit from AI referrals	53% higher than other shoppers
Consecutive months AI traffic outperformed on conversion	11

What that means: visitors arriving from AI aren’t browsing. They’re buying.

That tracks with how the journey works. The user was in the middle of a conversation about what they needed when they got sent to your page. They arrive with an intent that’s already formed, not a vague keyword.

So the first conclusion: product pages weren’t replaced. They got a new upstream entrance.

Two: the premise that “AI only reads data” is wrong

The question assumes AI learns about your products from feeds and schema.

That assumption is only half true, and it has three holes in it.

Hole one: schema is often wrong or incomplete

Schema is a claim, not a fact. Fields go missing, types get written wrong, and the markup drifts out of sync with what’s actually on the page.

The moment an AI system decides your schema isn’t trustworthy, it has to go somewhere else to cross-validate. That somewhere is most likely your product page body copy — because that’s the only original description of your product you fully control and that can be read.

Hole two: feed fields can’t hold the details that actually close a sale

Feeds come with format constraints: character limits, image size requirements, different mandatory fields per platform. You cannot put these in a feed:

• How to pick a size — “runs tight, size up half a size”;
• Compatibility — “fits an 11-inch iPad Pro, not the 12.9”;
• Real use cases — “built for flat feet on trails,” not “great for sports”;
• Honest battery numbers — lab figures versus what buyers actually get;
• The edges of your returns and warranty policy.

These are exactly the details AI needs to answer “will this work for me?” There’s nowhere to put them in a feed. There is plenty of room on the page.

Hole three: retailers that lean on feeds are widely invisible

In July 2026, SALT.agency audited product pages from leading brands and found that 70% don’t currently meet Google’s Universal Commerce Protocol (UCP) requirements — which makes them invisible in agentic shopping.

That statistic is its own rebuttal: when 70% of an industry isn’t ready, “a feed is enough” isn’t reality. It’s a hope.

Three: the product page’s real new job is being the source of truth

I’d sum up the new role in three words: source of truth.

Here’s the chain:

Agents read your schema → schema is incomplete → they need to cross-validate → third-party content becomes a source → and third parties learned what to say by reading your product page.

Let me unpack that.

1. Your product page teaches the whole ecosystem how to describe you

There’s a group of people you can’t be online to answer for: influencers, affiliates, ecommerce editors at media companies, creators you’ve onboarded into your program.

Before they write anything, they research your product pages. If the page says “built for overpronators on technical trails,” that’s what they’ll write. If the page says “high-quality athletic shoe,” that’s what they’ll write too.

When every trusted source says the same thing, AI has a reasonable basis to conclude that this product and this page are the ones worth recommending.

Flip it around: a selling point that never made it onto the product page is one fewer thing for the entire ecosystem to learn.

2. Product pages are how crawlers discover your other pages

Links in feeds get crawled. Crawlers follow internal links from your product pages to discover new models, alternatives, use-case pages, and comparison pages.

Skip the internal links and you’ve blocked the aisles in your own store — crawlers only see the handful of SKUs you submitted.

That matters more in the AI era, because competition in the citation pool is really competition over how many of your pages are indexed at all.

3. Product pages are trust devices

The Ask an SEO author told a very familiar story: he researched a jacket on the brand’s own site and watched YouTube reviews, then bought it on Amazon instead. Why? The brand’s site had no clear return policy; Amazon did. The product page never earned enough trust to close on its own.

He also described a comparison round: of five companies, three eliminated themselves on product page experience alone — no details, no humanity, felt like knock-off sites — leaving two finalists.

Note the order: the product page decides whether you’re even in consideration. Price and shipping come after.

Four: don’t mix up your two sets of copy — cannibalization is real

Most brands have already stepped in this trap, so it deserves its own section.

Many brands maintain a second set of product descriptions for marketplaces, affiliates, and third-party platforms — specifically to keep those channels from outranking their own product and category pages.

The reason is practical: search engines and AI systems often can’t tell who the original author or content owner is. Hand over your best copy as-is and a platform with more domain authority can end up ranking for it — meaning the branded searches you paid to create land on someone else’s page.

So the strategy inverts:

• Keep the fullest version on your own site — every detail, use case, data point, comparison;
• Distribute a leaner version externally — enough to satisfy mandatory platform fields, no exclusive information;
• The goal is for your page to be the one AI recommends, not for your copy to get someone else recommended.

In one line: your copy distribution policy is the moat around your source of truth.

Five: what an AI-era product page should contain

This checklist can go straight into your template.

Content	Why it matters (for AI and humans alike)	Example
Specs and compatibility	AI needs it to answer “will this work?”	“Fits an 11-inch iPad Pro,” with a size chart
Sizing logic	Cuts returns; AI repeats it	“Runs tight — size up half; wide feet stay true”
Use cases and audience	Lets AI judge “is this for me?”	“For flat feet on trails; not for commuting”
Comparisons and alternatives	Captures “which is better than X” questions	Spec table plus who each one suits
Returns, warranty, shipping	The trust device that leads to add-to-cart	Days, conditions, who pays return shipping
Data and provenance	Gives AI something citable	Test results, test conditions, standard numbers

Plus three writing habits:

1. Clear, readable heading structure — AI reads the structure, humans read the structure. No conflict;
2. The page has to render properly — botched JavaScript means AI reads a blank page, which is the real reason plenty of sites have clean feeds and zero citations;
3. Q&A-style short sections — turn the questions customers actually ask into subheads instead of stacking adjectives.

Six: Neo’s take — don’t move budget from product pages to GEO

I’ve seen this decision several times now: “AI is the trend, so let’s shift traffic budget into GEO. The old product copy can wait.”

The direction is right and the move is wrong. Three reasons.

One, product pages are where conversion starts, not an appendix to content. Every channel — AI referrals, organic search, email, SMS, affiliates, YouTube, social, paid ads — lands on a product page. If the product page doesn’t sell, all that spend just educated the market about your competitors.

Two, AI is one channel, not the channel — and it can change its mind about you at any time. No company survives on “being recommended by AI.” Today it recommends you, tomorrow it recommends someone else. Your email list, your organic rankings, and your repeat customers don’t vanish overnight — but they all point at the same product page.

Three, the value of being the source of truth goes up over time. The more pages AI crawls and the more often it cross-validates, the more weight “the only complete source” carries. This isn’t a one-off cost. It’s an asset.

So the order I’d give an independent site:

1. Fill in the product pages first (the six categories above), especially returns and compatibility, the two most commonly missing;
2. Then align schema and feeds with what the page actually says — not “write prettier schema,” but make the claims match the facts;
3. Only then work on distribution — full version on your site, lean version everywhere else;
4. Add the internal links while you’re in there, so crawlers can walk from product pages to your new models and use-case pages.

The end

I agree with the closing line of that column, roughly: if you don’t update your product pages, your competitors will — and they’ll pass you because they were thinking about their customers rather than about AI systems.

That gets to the root of it: the product page in the AI era isn’t written for AI. It’s written for people. It’s just that AI now reads it first on the customer’s behalf — and pages that are clear get one more shot at being recommended.

Don’t starve the one page every channel converts through, just to chase a new one.