AI content in e-commerce: why it costs you rankings in 2026

AI content is now costing e-commerce brands measurable rankings

AI-written product content rarely ends up on top in Google. In a Semrush analysis of 42,000 blog posts from April 2026, the text at position 1 was classified as human-written 80% of the time and as purely AI 9% of the time. The gap is widest at the top: at position 1, a page is eight times more likely to be human-written than AI-written.

That is not a punishment for using AI. Google says it judges what a page offers the reader, not how it came about. But pure AI text lacks exactly what Google does reward: original information, first-hand experience, a reason to exist. And anyone generating pages at scale without added value falls under Google's spam policy against scaled content abuse, whether a person or a model wrote the text.

This article explains what Google did and didn't change in 2026, how to tell whether your webshop is at risk, and what you concretely do to protect your rankings.

What exactly changed with the March 2026 Core Update?

Less than many reports suggested. The broad core update of March 2026 ran from 27 March to 8 April and Google called it a regular update, with no companion blog post and no announced goal. There was no new rule against AI content.

The rules that do matter have been in place for longer. In March 2024 the old helpful content system became part of Google's core ranking systems and Google introduced three new spam policies, including one against scaled content abuse: creating many pages to manipulate rankings rather than help people, "no matter how it's created". Google's own guidance on generative AI content puts it even more directly: AI is useful for research and structure, but generating many pages without value for users may violate the spam policy.

The core remains E-E-A-T: Experience, Expertise, Authoritativeness, Trustworthiness. Pure AI text falls short mainly on the E of Experience: a model cannot claim hands-on experience with your product.

What Google actually looks at

Google publishes a list of questions you can use to assess your own content. For product pages they come down to five signals:

• Original information. Google literally asks whether content provides "original information, reporting, research, or analysis". Generic phrasing that fits any product does not.
• Demonstrable expertise. Is the text written or reviewed by someone who demonstrably knows the subject?
• Clear sourcing. Claims that call for evidence get a source.
• A visible creator. Who made the content, and can that be checked through an author page or about page?
• Care per page. Google names content that is mass-produced, so that individual pages get little attention, as a warning sign.

This does not mean Google detects and punishes AI content as such. It means content without expertise is less valuable to users, and that AI makes it easier to produce such content at scale.

The timeline: what actually happened

The lesson is not "don't use AI." The lesson is: AI without expertise produces content without value, and that rarely ends up on top.

Why does this hit e-commerce the hardest?

E-commerce is vulnerable because the temptation to use AI at scale is greatest there and the texts naturally resemble each other. There are three concrete reasons.

Reason 1: Scale made it tempting. A webshop with 500 products needs 500 product descriptions. Plus category pages. Plus blog content. AI made it possible to produce that in days instead of months. But exactly that pattern, many pages with little original substance, is what Google describes as scaled content abuse.

Reason 2: Product content is inherently similar. When ten competitors use the same AI tool to generate a description for the same product category, you get ten variations on the same generic text. Google sees ten pages that offer nothing unique. None of those pages earns a top position.

Reason 3: E-commerce content needs specific expertise. A good product page contains information only someone with real product experience can write. How does the material feel? Which size runs large? What do customers use it for? What do buyers ask most often? AI cannot make this up. It has to come from people who know the product.

We see this often with brands that come to Oase Creative for SEO help. The product descriptions read as if they were written by no one in particular. There is no voice, no expertise, no reason why you should be the source.

The parallels with earlier updates

This pattern is not new. In 2011 Google's Panda update hit content farms that published thin articles en masse. In 2012 Penguin tackled link-building abuse. In 2022 the Helpful Content Update targeted content written for search engines rather than for people, and in 2024 that idea was absorbed into the core system.

Every time the story was the same: a shortcut that worked for a while, until Google raised the bar. AI content at scale is the latest version of that story.

How do you tell whether your webshop is at risk of losing rankings to AI content?

You are at risk when your product pages are interchangeable: text that could sit on a competitor's site without changes. Before you act, you need to know how big the problem is. Here are the five signals that your product content is vulnerable.

Signal 1: Generic product descriptions. Open five random product pages in your webshop. Read the descriptions. Could you guess which product it is without the product name and photo? If the description would fit ten other products in the same category, it is generic.

Signal 2: No FAQ sections with real customer questions. Product pages without FAQs miss one of the strongest signals for both Google and AI search engines. Real FAQs are based on questions customers actually ask, not on what an AI tool generates as "frequently asked questions".

Signal 3: No comparisons or specifications. Shoppers look for comparisons. "Product A versus product B." "Which size should I take?" If your product pages contain no comparative information, you miss both search traffic and AI citations.

Signal 4: Traffic decline since March 2026. Check Search Console. Compare April-September 2026 with the same period in 2025, per page type. A clear organic decline on product or category pages while your brand name holds steady is a signal to review the content of those pages.

Signal 5: The same content structure on every page. Open your product pages side by side. If every page follows exactly the same layout (intro, three features, call to action) without variation, it looks like template content. Templates are not bad by definition, but templates without unique substance are.

The quick audit

Run this check on your top 20 product pages (ranked by revenue):

Three or more "No" answers on your top pages? Then action is needed. Now, not next month.

What makes hybrid content better than pure AI content for SEO product pages?

Hybrid content combines the speed of AI with the product knowledge of a person, and only that combination is both scalable and original. Pure human content is good but does not scale. Pure AI content scales but lacks the original substance that ends up on top.

Hybrid content means: using AI as a tool, but putting final responsibility with someone who knows the product and the market. The AI sets up the structure, the human fills in the expertise.

Why hybrid wins

It is also how most teams already work. In Semrush's April 2026 survey, 64% of SEO teams worked with a human in charge and AI assisting, and 23% created content entirely without AI. Semrush describes its own process the same way. Ana Camarena, Head of Organic Content Strategy at Semrush, writes in that study: "we have not replaced editorial judgment with AI, and we are not rushing to do so."

Note what that study does not say. The AI detector looked at the finished text, not at the process. Text prepared with AI and rewritten by a skilled person reads like human work. That is exactly where the gain of working hybrid lies.

What hybrid content looks like in practice

A hybrid product page for a running shoe:

What the AI does:

• Set up the structure (specifications, comparison, FAQ)
• Gather basic information (material, weight, drop)
• Generate a first version of the description

What the human adds:

• How the shoe actually feels after 10 km
• Which foot type it suits best (based on customer feedback)
• A comparison with the previous model based on own experience
• Answers to questions customers ask in reviews and via customer service
• Photos of the product in use

That human layer is exactly what feeds Google's E-E-A-T signals. It is the difference between a product page that informs and one that converts.

At Oase Creative we use this hybrid model for every Shopify webshop we build. AI helps with speed. The expertise comes from our team and from the client. Where AI stops when building the shop itself is covered in building a webshop with AI: the last twenty percent.

How Google measures E-E-A-T on product pages

Google does not measure E-E-A-T with a single score, but with signals that together show who is behind a page and why you can trust it. Product pages involve money people spend, so trust carries a lot of weight.

Experience: proof of first-hand experience

Google looks for signals that the author or brand has actually used or worked with the product.

Concrete signals:

• Your own product photos (not just supplier photos)
• Descriptions with sensory detail ("the material feels", "the scent is")
• Comparisons based on own use ("compared with the previous model")
• Customer cases with specific results

Expertise: proof of subject knowledge

Google assesses whether the content comes from someone with relevant knowledge.

Concrete signals:

• Author information with a verifiable background
• Detailed technical specifications with explanation
• Contextual recommendations ("for beginners we recommend X, for advanced users Y")
• Sources for claims

Authoritativeness: proof of recognition

Google measures whether others see you as an authority.

Concrete signals:

• Backlinks from relevant sources
• Mentions on industry sites
• Reviews and ratings on your own site and on external platforms
• Consistently publishing on the same topic

Trustworthiness: proof of reliability

Google checks whether your site is trustworthy for transactions.

Concrete signals:

• SSL certificate and secure checkout
• Clear returns policy and contact information
• Customer reviews on the site and on external platforms
• Transparent pricing information

Top-ranking e-commerce content has these four elements in order consistently. Not as an SEO trick, but as a fundamental part of how the site is built.

What AI Overviews mean for your product content

AI Overviews mostly appear for informational questions and far less for buying questions, but the share of buying questions is growing. The Semrush AI Overviews study shows that transactional queries triggering an AI Overview grew from 1.98% (October 2024) to 13.94% (October 2025). The Shopping category actually lags within that, with the smallest share and little growth.

The categories hit hardest

Growth is not evenly spread. In the same study, Real Estate and Shopping were among the sectors with the smallest share of AI Overviews, while categories such as food and drink rose sharply. The pattern is clear: no category stays out of range in the long run, some are simply up sooner than others.

Where AI Overviews do appear, the impact on clickthrough rate is large. Ahrefs measured that, for keywords with an AI Overview, the number 1 position had a 58% lower clickthrough rate in December 2025 than would have been expected without the AI Overview. The traffic that does come through is valuable: according to Adobe Analytics, visitors from AI sources converted 54% better on US retail sites in May 2026 than visitors from other sources.

What this means for your strategy

The takeaway is not panic but precision. Product pages are relatively safe for now. Informational content (blogs, guides, explainers) is being summarised en masse by AI Overviews.

This changes your content strategy in two ways:

1. Product content: make it worth citing. When AI Overviews reach your category (and they will), you want to be the source Google cites. That takes structured data, concrete specifications and unique product insight. Exactly the things pure AI content lacks.

2. Blog content: put your information up front. Kevin Indig found in a large-scale analysis of ChatGPT answers that 44.2% of citations come from the first 30% of the text. If your key point sits on page two, it gets cited less often. Put the answer up front.

Which concrete steps protect your rankings now?

Start with your best-selling pages, make them demonstrably your own, and measure the difference. Here is what you concretely do, ranked by impact and feasibility.

Step 1: audit your current content (day 1-3)

Pull a list of your top 50 product pages by revenue and search volume. Score each page on the checks from the audit table earlier in this article. Mark pages with three or more "No" answers as a priority.

Step 2: rewrite your top 20 product descriptions (week 1-2)

Start with the twenty pages that generate the most traffic and revenue. Add to every page:

• Real use cases based on customer feedback and own experience
• Specific product knowledge that is not on the manufacturer's site
• An FAQ section with at least four questions customers actually ask
• A comparison with at least one alternative product
• Structured data (Product, AggregateRating, and FAQPage where you have FAQs)

Step 3: implement structured data sitewide (week 2-3)

Structured data is the easiest technical win. Google and AI search engines use schema markup to understand and cite product information.

The minimum per product page:

• Product schema with name, description, price, availability
• AggregateRating schema with average score and number of reviews
• FAQPage schema for FAQ sections (Google no longer shows a rich result for it since May 2026, but the question-and-answer structure stays usable for AI systems)
• BreadcrumbList for navigation structure

Using AI for titles and descriptions in your product feed? Then according to Google Merchant Center those belong in the structured_title and structured_description attributes, with trained_algorithmic_media as the source. Product images made with AI must keep their IPTC metadata.

Step 4: build an expertise layer into your content (ongoing)

This is the structural change. Stop publishing content anyone could have written. Start with content only you can write.

Concrete actions:

• Link every product page to an author with verifiable expertise
• Add customer cases where possible (anonymised is fine)
• Publish comparative content based on your own tests
• Answer product questions in depth, not with generic replies

Step 5: monitor and iterate (monthly)

Measure the impact of your changes in Google Search Console:

• Compare impressions and clicks per product page before and after the rewrite
• Check in the generative AI report (in Search Console since 2026) which pages appear in AI Overviews and AI Mode
• Track which pages recover and which do not
• Identify patterns: which type of content works best in your category?

Bol.com proves it: small SEO tweaks deliver measurable revenue

Small, targeted improvements to product content deliver measurable results, even at the largest player in the Netherlands. At Friends of Search 2026, bol presented an experiment in which product titles and H1 headings were improved with AI and customer input. The result: a 3 to 4% increase in clicks and GMV (Emerce, March 2026).

That is bol. A platform with millions of products and an established SEO position. If targeted improvements already deliver clicks and revenue there, the room for a webshop that has never worked seriously on product content is larger.

The difference lies in the method. Bol used AI as a tool to improve existing titles, fed by what customers themselves indicated, not as a replacement for knowledge. Hybrid, not pure AI.

This confirms the pattern: AI as a tool works. AI as a replacement for expertise does not.

What you can learn from Bol.com

• Titles and H1s are the easiest win. Start there. Not with rewriting your entire site.
• Test at small scale. Take ten products, optimise the titles, measure the difference over four weeks.
• Use AI for wording, not for substance. The expertise is already in your head or in your customer data. AI helps you write it down better.

Reviews: why social proof measurably counts in AI recommendations

Reviews are the signal AI systems can compare most easily, and a small difference in rating can tip the recommendation. In an experiment we discuss in our piece on brand preference in AI recommendations, the advantage of a well-known brand name almost completely disappeared as soon as a small difference in rating was visible.

That makes sense. AI search engines want to cite reliable sources and recommend products others have already approved. A product with a handful of reviews is an unknown. A product with a long, recent run of reviews is proven.

How to build review counts

• Automate review requests. Send an email 7-14 days after delivery with a direct link to your review page.
• Make it easy. One click, no account needed, no long forms.
• Respond to reviews. Customers see that you care about feedback, and your replies add substance themselves.
• Use reviews in your content. Quote frequently mentioned pros and objections on your product page. This adds real user experience to your content.

There is no hard minimum number of reviews above which AI suddenly recommends you. Start with your best-selling products and work step by step towards more and more recent reviews. Every step improves your position against a competitor with less proof.

The technical side: why page speed now shapes rankings too

Speed is a separate signal next to content: the best product content does not help if the page loads slowly or jumps around. Google measures that with the Core Web Vitals, and the thresholds for a good score are on web.dev.

Core Web Vitals you have to measure

Quick wins for e-commerce sites

• Optimise product images. WebP format, lazy loading below the fold, correct dimensions. This is one of the most common causes of slow LCP on product pages.
• Minimise JavaScript on product pages. Every tracking pixel, widget and popup slows down INP. Measure what you really need.
• Use a CDN. Static assets belong on a CDN, not on your origin server. This improves load times everywhere.

Technical SEO and content SEO reinforce each other. The best product content in the world does not help if your page needs five seconds to load. And the fastest site in the world does not rank if the content is generic.

What Oase Creative does differently

We combine SEO, product knowledge and our own imagery under one roof, so a product page gets substance a competitor cannot copy. We have been building brands for six years and have worked for more than 200 brands in that time. What Google showed in 2026 confirms what we already knew: shortcuts do not work in the long run.

At Oase Creative we combine SEO expertise with real product knowledge. Our approach:

Content that can only come from you. Every product description we write contains information your competitor does not have. Your own product photos, your own customer insights, your own use cases.

Technical SEO as the foundation. Structured data, Core Web Vitals, internal link structure. The basics have to be right before you start improving content.

Hybrid workflow. We use AI as a tool, not as a replacement. AI helps with structure and speed. The expertise comes from our team and from you as the brand owner.

Everything under one roof. SEO, Shopify development, product photography, branding. That combination is exactly what e-commerce brands need to deliver the E-E-A-T signals Google looks for. Not five separate freelancers each doing a piece, but one team that sees the whole picture.

How to build a content strategy that survives future updates

A content strategy that survives updates revolves around substance only you can deliver, structured so people and machines can use it. The March 2026 core update is not the last. AI Overviews and AI Mode will absorb more search traffic. The only strategy that works in the long run is content that genuinely adds value.

The framework in three principles

Principle 1: Write from experience, not from templates. Every product page has to contain information only you have. That is the definition of expertise in the context of E-E-A-T.

Principle 2: Make your content citable. Structure your information so AI systems can extract it. Concrete facts, tables, FAQs, sources. The easier it is to cite, the more often it gets cited.

Principle 3: Invest in proof. Reviews, cases, results. Not telling people your product is good, but showing that others think so.

The monthly check

Build a monthly routine:

• Check Google Search Console for changes in impressions and clicks per product category
• Compare your top positions with three months ago
• Identify declining pages and analyse why
• Rewrite at least five product pages per month, starting with the most important
• Add structured data to pages that still lack it

Consistency beats speed. Rewriting five pages a month properly is better than having AI generate five hundred in a weekend.

The bottom line

Pure AI content rarely ends up on top, and that is down to what is missing from it, not to the tool. Google's March 2026 core update was no reckoning with AI. It is the logical next step in a pattern that has been visible for years: shortcuts lose, expertise wins.

In Semrush's measurement, 80% of position 1 results read as human-written, against 9% purely AI. Anyone generating worthless pages at scale also runs into Google's spam policy.

The solution is not to stop using AI. The solution is to stop using AI as a replacement for expertise. Use AI for structure, speed and scale. But make sure every page contains information only you can provide.

Start with your top 20 product pages. Add real product knowledge, FAQs based on real customer questions, comparisons with alternatives, and structured data. Measure the difference. Scale what works.

The brands doing this now are building a lead that will be hard to close later. Just like the early SEO adopters profited from their Google positions for years.

Need help setting up a content strategy that holds up against algorithm updates? Get in touch for a conversation about what we can do for your brand.