Almost everything written about product page optimisation assumes one reader: a person, scrolling, deciding. Write better descriptions, add more images, collect reviews, speed the page up. All of it still true, and all of it now half the job, because in 2026 your product page has to convince something else first, and that something decides whether the person ever sees you at all.
The shopping journey increasingly begins with an assistant that compares products, shortlists them, and, in a growing number of cases, completes the purchase. So here's the frame for this whole guide: your product page has to sell twice. Once to the machine, which is looking for facts it can verify. Once to the human, who is looking for a reason to trust you. Those aren't the same job, and the places where they conflict are exactly where e-commerce most likes to be clever.
Reader one: the machine that shortlists you
Before a human sees your page, a system may already have compared it against your competitors. It doesn't scroll. It doesn't admire your typography. It extracts, and if it can't extract, it moves on. Being excluded at this stage is silent, which is what makes it dangerous, because you'll never see the customer you never got.
What it needs is dull and non-negotiable: price, availability, delivery cost, delivery time, returns policy, and specifications, all as machine-readable text. Not implied. Not in an image. Not calculated at checkout.
Product schema is not optional here
An important nuance, because it's widely misreported. Google has said structured data isn't strictly required for its generative AI features, and that's true. But e-commerce is the great exception, because shopping surfaces and rich results read your price, availability, and review data directly from your markup. Product schema is how you get into a comparison at all. If your competitor's price and stock status are structured and yours aren't, they are comparable and you are not, and nothing that isn't comparable gets recommended.
The uncomfortable rule A fact a machine can't read is a fact you don't have. Your price, your stock level and your delivery promise either exist as data, or they effectively don't exist at all, however beautifully they're rendered for humans.
The clever tactics that now delete you
Here's where it stings. Several long-standing e-commerce tactics were designed to control the human experience, and in an AI-mediated journey they now cause exclusion.
| The tactic | Why you did it | What it costs now |
|---|---|---|
| "Add to cart to see price" | Protect margin, create curiosity, dodge price comparison | You're not in the comparison at all |
| Price or specs baked into an image | Design control, promotional flexibility | The facts are unreadable, so unverifiable |
| Shipping cost revealed only at checkout | A lower headline price | Total cost can't be computed, so a rival who shows it wins |
| Sizing guide as a PDF | It was easy | You can't be the answer to "will it fit a 40in chest?" |
| Specs in a tab that loads on click | A cleaner layout | Content may never be seen by anything that doesn't click |
Notice the pattern: every one of these was a conversion decision that has quietly become a visibility problem. It's the same structural issue we cover in designing websites for AI search, but for retailers the stakes are unusually literal, because the trapped facts are the ones that decide the sale.
Reader two: the human who actually lands
Now the person arrives, perhaps already half-sold by an assistant, perhaps still comparing. Your job changes completely. The machine wanted facts. The human wants certainty, and every unanswered question is a reason to leave.
The five questions, all answered without a click
Most product pages answer the first of these and bury the rest in tabs, accordions, or the checkout, which is where a large share of abandonment quietly originates.
1. What exactly is this? Materials, dimensions, what's in the box.
2. What will it cost me in total? Including delivery. A surprise cost at checkout is the most reliable way to lose a sale you'd already won.
3. Is it actually in stock? In my size, my colour, right now.
4. When will it arrive? A date, not "3–5 working days" from an unstated starting point.
5. What if it's wrong? Returns, cost, and window, stated plainly and without shame.
That last one is worth defending, because retailers hide returns policies fearing they'll encourage returns. In practice a clear, generous policy reduces hesitation and lifts conversion, because it removes the risk the customer is silently weighing. Hiding it doesn't stop returns. It stops purchases.
Images are the product
Online, images do the job that touching the product does in a shop, which makes a thin gallery an unforced error. You need every angle, close-ups showing texture and finish, something that conveys real scale (the object held, worn, or beside something familiar), the product genuinely in use, and, best of all, customer photos.
Real customer images outperform studio shots for one simple reason: shoppers trust them more. They show the product as it actually arrives, in real light, on real people, which is the same authenticity dynamic driving everyday creators outperforming celebrity endorsements. Usefully, this is one of the cases where both readers benefit at once, since images with real alt text are legible to machines too.
Descriptions should kill objections, not stack adjectives
Most product copy is a pile of adjectives written by someone who already believes in the product. "Luxurious." "Versatile." "Premium." They persuade nobody, because they answer no question anyone was actually asking.
Write against the objection instead. Go to your customer service inbox, find the questions people ask before they buy, and answer them on the page. Does it run small? Is it noisy? Will it work with the older model? Is the fabric scratchy? Those answers convert, and conveniently they're also the specifics that make you retrievable, because they're the phrases real people search. Generic AI-written descriptions fail on both fronts simultaneously, which is the content sameness problem arriving on your storefront.
Reviews are the closing argument
Reviews are the most powerful element on a product page and the most under-managed. Some hard-won principles: volume matters more than perfection, and a wall of unbroken five-star reviews reads as fake. A few critical reviews with a good-faith reply from you build more trust than a spotless page. Answer the negative ones publicly. And surface the reviews that address common objections, not merely the most enthusiastic ones.
Where the two readers agree
Encouragingly, the overlap is large, and it's where you should spend first.
| What you do | Why the machine likes it | Why the human likes it |
|---|---|---|
| Real specifics, not adjectives | Extractable facts to match against a query | Answers the question they're actually asking |
| Everything as text, not images or PDFs | Readable, therefore comparable | Accessible, searchable, works on any device |
| Total cost and delivery date up front | Can compute and compare the real price | No nasty surprise at checkout |
| Genuine, plentiful reviews | Evidence for subjective "best X" queries | Proof from people like them |
| A fast, uncluttered page | Crawlable and cheap to fetch | Doesn't punish them for arriving on mobile |
The technical floor
Unglamorous, and it will quietly cap everything above it if you get it wrong.
- Clean, readable URLs that describe the product rather than a string of parameters. They're for people first; stuffing them with keywords helps nobody.
- Handle variants deliberately. Twelve colourways spawning twelve near-identical pages is a duplication problem. Decide, with canonicals, which version you want to rank, and make sure the variant a customer needs is still reachable.
- Never let out-of-stock pages 404. Keep the page, state the status honestly, offer alternatives or a restock alert. You worked hard for those links; don't delete them because a warehouse ran dry.
- Speed, seriously. Product pages are where heavy scripts accumulate: review widgets, upsell apps, chat, personalisation. Audit what each one actually earns you.
- Get the fundamentals right first, per our on-page SEO checklist. Nothing exotic beats a page that's fast, semantic, and honest.
Test the way you're actually being read
Two diagnostics worth running this week, both free.
First, ask an AI assistant to compare your product against two competitors. See whether it knows your price. See whether it gets your delivery time right. See whether it mentions you at all. That's a more honest read on your product page's health than any rank tracker, and it's the same test we recommend for AI-driven local discovery.
Second, disable JavaScript and reload your product page. Whatever disappears is at risk. If the price, the specs, or the reviews vanish, you've just watched your page through the eyes of the thing deciding whether to recommend you. And if PDP traffic is sliding while rankings hold, resist the panic-redesign and diagnose it properly with an end-to-end funnel audit first. If the fix does turn out to be structural, that's exactly what e-commerce marketing work is for.
The bottom line
Product page optimisation used to mean persuading a person. It still does, but only after you've satisfied the system that decides whether that person ever arrives. So publish your facts as facts, price, stock, total cost, delivery date, returns, in plain readable text, because anything trapped in an image, a PDF, or a click-to-open tab is a fact you effectively don't have. Then win the human with what machines can't fake: photographs that answer doubt, copy that kills objections instead of stacking adjectives, and honest reviews you haven't sanitised. The two jobs sound opposed and mostly aren't. Specificity, completeness and honesty serve both readers, while every clever tactic built on withholding information now serves neither. In 2026, the merchant with nothing to hide is the one that gets recommended.
Are your product pages getting skipped?
Make your product data readable to AI and persuasive to shoppers, so you win both.
Explore E-commerce Marketing →Frequently asked questions
Do I still need product schema in 2026?
Yes, and e-commerce is the clearest case. Google has said structured data isn't strictly required for generative AI features, but shopping surfaces and rich results read price, availability, and reviews directly from your markup. If a competitor's price and stock are structured and yours aren't, they're comparable and you're not.
Why does hiding prices hurt product pages?
Because a price that only appears in the cart, or is rendered as an image, can't be read by the systems now comparing products on a shopper's behalf. "Add to cart to see price" was designed to protect margin or create curiosity. Today it usually means exclusion from the shortlist before a human ever sees you.
What should a product page answer above the fold?
Five things: what it is, what it costs in total including delivery, whether it's in stock, when it will arrive, and what happens if it's wrong. Most pages answer the first and hide the other four behind tabs or the checkout, which is where most abandonment quietly begins.
How many images should a product page have?
More, and of more types, than you have now. Cover every angle, close-ups showing texture and finish, something conveying real scale, the product genuinely in use, and customer photos where possible. Images do the job of touching the product in a shop, and thin galleries cause hesitation at the worst moment.
