The half-life of a product page.

Consider two product pages, launched on the same Monday, for the same kind of product, by two different stores.

The first page is built carefully. A copywriter has been paid to write a hero paragraph. The marketing team has chosen three benefits, two ingredients to call out, one founder story. The page ships with a hundred-and-fifty-word product description, a feature grid, a frequently-asked-questions block written by the brand, and a review widget under the buy button. The page is reviewed by the brand's founders, lightly edited, approved.

The second page is built less carefully. The hero paragraph is shorter. The feature grid is replaced by a paragraph the founder wrote on a Saturday morning. There is no FAQ block of the marketing team's own design. There is, however, the same review widget under the buy button.

A year later, the first page looks almost identical to the day it launched. The hero paragraph is intact. The feature grid is intact. Two of the three benefits have been reworded. The review widget under the buy button now displays four-point-six stars instead of four-point-two. The product description has been updated once, in March, when the supplier changed.

A year later, the second page is unrecognisable. Not because the brand redesigned it, but because the brand structured the page so that every review written about the product became part of the page itself, indexed, answerable, surfaced in the page's own prose, and not buried in a JavaScript carousel.

The first page is, in the language of organic search, decaying. The second page is compounding. The conventional reading of product page longevity SEO treats the page as a thing to refresh on a calendar; the reading we are about to make treats the page as a thing the customers refresh on the brand's behalf.

A product page is not a thing you ship. It is a thing you accept the responsibility for, then leave alone while the people who buy from you write it for you.

This essay is about product page longevity SEO, by which we mean the specific property of a page that allows it to gain organic traffic over the years it lives, rather than lose it. The argument is simple. The marketing copy on a product page decays the moment it is written. The customer-written language that the same product accumulates compounds for as long as the store stays open. A merchant who structures the page to absorb that language, in public, in plain HTML, owns an asset that very few competitors are bothering to build.

There is, conveniently, a name for what that asset gives the merchant in return. The trade discusses it under product page longevity SEO, when it discusses it at all; in practice the term is doing two different jobs at once, and only one of them is interesting. The first is the calendar refresh, which we will dispense with quickly. The second is the slow accumulation of customer language, which we will spend the rest of the essay on.

There is a piece of folk wisdom in content marketing that says a blog post has a half-life of about two years. After two years, the post is sending half as much traffic as it sent at peak, because the world has moved on and the search engines have moved on with it. The wisdom is wrong in the specifics. It is right in the shape.

Marketing copy decays. A 2024 blog post titled "the best way to ask for reviews on Shopify" was written for a search landscape in which Google delivered ten blue links to a user who would scan them. The same post in 2026 is competing with a Google AI Overview that answers the question above the fold, an answer-engine summary in ChatGPT, and a Reddit thread that ranks higher than the post does for the same query. The post has not changed. The world the post was written into has changed. The post is now worth a third of what it was. The half-life is real.

Marketing copy on a product page decays in a slightly different way. It does not lose to a Google update; it loses to its own staleness. The hero paragraph the brand wrote in 2024 said the product was for "modern, busy people". The hero paragraph in 2026 still says that. The hero paragraph in 2028 will still say that. It is wallpaper. It is decorative. It tells the answer engine, and the buyer, very little.

The brand can refresh the paragraph. Most brands do. They sit down twice a year, decide the copy feels tired, hire a copywriter, change "modern, busy people" to "people who actually have lives". The half-life is reset for a quarter; the page picks up a small amount of new traffic; the page returns to wallpaper.

This is what the half-life of a product page looks like under the standard model: a sawtooth, sloping down. Every refresh buys a month of growth. Every month between refreshes loses two months of relevance.

The total work involved, over a year, is considerable. The total traffic gain, on a per-product basis, is approximately nothing.

Now look at the second page.

In its first month, the page receives twelve reviews. The reviews are short, mostly four and five stars, with a few sentences each. One reviewer mentions that the product worked under SPF. Another mentions that they have eczema and the product did not irritate it. A third mentions that they bought it for their teenager.

None of those facts are in the marketing copy. None of them would have been; the marketing team had no way of knowing what the buyers would mention. The marketing team is paid to imagine the buyer; the buyer is the actual buyer.

In its first month, the page is structurally identical to the first page. Same hero paragraph, same feature grid, same review widget. The reviews are sitting in the widget. They are visible to anyone who scrolls down. They are invisible to almost everyone else.

In its second month, the brand's review system does something different. It does not bury the new reviews in a paginated carousel; it publishes them as text on the page. It writes a short, signed, dated reply under each review, in the brand's voice, in public, in indexable HTML. It generates a small Q&A block on the page from the questions buyers have asked the support team about this product. It updates the page's structured data to include the new reviews as `Review` entities under the product, with body text and authors and dates.

By the third month, the page has, in addition to the marketing copy, a body of customer-written language that the marketing team could not have produced. The sentence "did not irritate my eczema" is now present on the page. The phrase "held up under SPF" is now present on the page. The note about buying it for a teenager is now present on the page. They are present in plain HTML, in prose paragraphs, indexed by the engines that read the page.

By the sixth month, the page has accumulated forty or fifty such phrases. By the ninth, a hundred. By the twelfth, two hundred and more.

A product page that absorbs its own reviews is a page that gets longer, more specific, and more useful every week, without the brand writing a single new sentence.

The half-life of this page is no longer a half-life. The page is not decaying. The page is, by the simple measure of indexable specific language about the product, growing.

There is a useful number to hold in mind. According to BrightEdge's holiday 2025 audit, Google's AI Overviews cite a retailer's own pages only about four percent of the time when answering a product question. The remaining ninety-six percent comes from somewhere else: Reddit, Wirecutter, magazines, YouTube transcripts, forum threads, Trustpilot. The brand's product page, the one the brand has been paying a copywriter to refresh, is almost invisible to the engine that decides what the buyer reads.

The reason is not malice on Google's part. The reason is mechanical. The Princeton and IIT Delhi paper on Generative Engine Optimization, from 2024, measured what kinds of text the answer engines lift into their answers. The text was first-person, dated, entity-dense, and source-able. The text contained statistics, quoted language, citations. Pages with those features were surfaced up to about forty percent more often in generative answers than identical pages without them. We have argued the engine-side of this case more fully in the engine the answer engine reads.

The first product page has none of those features. The hero paragraph is third-person, brand-voice, undated, low-entity, citation-free. The review widget contains text that is first-person and dated and entity-dense, but the widget is a JavaScript component that AI crawlers either render unreliably or do not render at all.

The second product page is, by construction, the inverse. The page contains the marketing copy at the top, then a growing body of first-person, dated, signed, indexable reviewer language underneath. The page is, in machine terms, what the answer engine has been told to look for.

A year of compounding, on the second page, produces a page that is cited. A year of refreshing, on the first page, produces a page that is wallpaper.

There is a smaller, more pedestrian way the second page wins, and it is the long tail.

A copywriter writing a product page in advance has no way of knowing what people will type into a search box about that product. They guess. They write the page to rank for "best hydrating serum" or "vitamin C serum for sensitive skin" because those are the queries the marketing team thought of in a planning meeting. The page is optimised for the queries the brand imagined.

The buyers do not imagine. The buyers ask. They ask, over a year, several thousand variations of the same questions. They ask whether it works under SPF. They ask whether it stings. They ask whether the packaging is recyclable. They ask whether it is okay during pregnancy. They ask whether it lasted long enough for the price. They ask whether the dropper works left-handed.

A few of these questions show up in the brand's FAQ. Most of them do not. The reason most of them do not is that the brand does not yet know the questions exist. The buyers ask the questions, in their own language, in their reviews, in their support tickets, in their post-purchase emails.

A product page that absorbs that language, in public, becomes a page that ranks for those queries. The page that contains the sentence "did not irritate my eczema" begins to rank, modestly, for the long-tail query "[product] eczema". The page that contains "held up under SPF" begins to rank for "[product] under spf sweat". The page that contains "bought it for my teenager" begins to rank for "[product] teen skin".

None of these queries have meaningful search volume on their own. None of them would justify a marketing-team meeting. Together, on a single page, over a year, they account for a meaningful share of the page's organic traffic. They are the long tail. They were written by the customers. They cost the brand nothing.

This is the dynamic Yotpo's blog has, more than once, written around without naming directly. This is the dynamic that Judge.me's documentation gestures at when it says "more reviews mean more long-tail keywords". The mechanism, named plainly, is that customer language is the cheapest and most authentic long-tail keyword research a brand will ever own. We have written about this at greater length in the long tail begins inside the review. The work, named plainly, is to publish that language on the product page, in the page's own prose, in HTML the crawler will read.

The first page, the one with the carefully-written hero paragraph, is not failing. It is doing what it was built to do. The buyer who lands on it from a brand-name search or a paid ad has a reasonable experience. The page converts at the rate the brand expects.

The first page is failing only against the second page. The cost is not visible in the brand's dashboard. The cost is in the traffic the first page is not getting, and the citations the first page is not earning, and the long-tail queries the first page is not ranking for.

The cost compounds. By month six, the difference between the two pages is small but legible. By month twelve, the difference is large. By month twenty-four, the second page is generating multiples of the first page's organic traffic, without the brand writing a single additional paragraph of marketing copy.

The first page will at some point be re-written. The hero paragraph will be sharpened. The feature grid will be refreshed. The brand will declare a new visual identity and spend a quarter migrating to it. The half-life will be reset. The sawtooth will continue.

The second page will be left alone. It will be left alone because there is nothing to refresh. The page is being refreshed every week, in language better than the marketing team could have written, by the people who actually used the product.

The deepest and least understood property of a customer-written product page is that it gets better the less the brand touches it.

A few specifics, for the operator reading this on a Saturday morning between orders.

The first thing to do, if you have not already done it, is to look at your product page in source view. Open the page. View source. Press control-F and search for the body text of one of your reviews. If the body text appears in the HTML, you are most of the way there. If it does not appear, your reviews are being rendered by JavaScript after the page loads. The page that Google sees, the page that ChatGPT sees, the page that the Perplexity crawler sees, is a page without your reviews on it.

If your reviews are not in the HTML, the engineering work to put them there is, in most review platforms, the difference between a free plan and a paid one, or the difference between the default install and a "server-side rendering" toggle nobody at the brand has been told about. The toggle is worth more than any blog post the brand will publish this quarter.

The second thing to do is to look at your structured data. The page should contain a `Product` schema entity, with `aggregateRating` and a `review` array. The review array should contain the full text of the reviews, with `author`, `datePublished`, and `reviewBody` fields populated. Most review widgets ship with structured data; many of them ship with structured data that points at the widget's iframe rather than the merchant's domain. This is worse than nothing.

The third thing is to look at what the page does with the language inside the reviews. A page that displays the reviews in a paginated widget is doing the minimum. A page that surfaces the most common questions from those reviews into a Q&A block, in HTML, is doing more. A page that publishes a signed brand-voice reply under each review, in HTML, is doing the most. The third option is the one that compounds.

There is no off-the-shelf product, as of this writing, that does all three of these things by default on a Shopify store. Most of the widgets do the first; some do the second; almost none do the third in a way that puts the brand's reply in indexable HTML on the merchant's own domain. The shape of the right tool, which we have argued for at length in the end of the review widget, is a system that writes the page beneath the widget. It is not yet a category.

A skeptical reader, at this point in the essay, would be entitled to say: this argument is not new. Search engine optimisation has been about long-tail keywords and structured data for fifteen years. The advice to publish your reviews as text is twelve years old. Why now?

There are two reasons why this argument is sharper in 2026 than it was in 2024.

The first reason is the answer engines. Google's AI Overviews launched generally in 2024. ChatGPT search launched at the end of 2024 and crossed roughly nine hundred million weekly active users by early 2026. Perplexity grew from a curiosity to a meaningful share of high-intent commercial queries. Adobe Analytics reported that traffic to U.S. retail sites from generative AI sources rose roughly twelve hundred percent in twelve months. The market shifted to a model where the page is read by a machine that summarises it into a paragraph for the buyer. The page that the machine prefers to summarise is the one with the first-person, dated, specific language inside it. That is the page that absorbs its reviews. That is the page that compounds.

The second reason is the changes to Google's organic ranking signals over 2024 and 2025. The Helpful Content signals, the December 2025 core update, the suppression of refreshed-evergreen patterns and thin AI-templated content: each of these tightenings rewarded pages with original, source-able, first-person material and penalised pages that were obviously written for the engine. The product page that absorbs its customers' language has, by construction, the kind of material the algorithm has been told to reward. The product page that ships a refreshed hero paragraph every quarter does not.

The two reasons stack. Pages with compounding customer language win in both the classic Google search and the new answer engines. Pages with refreshed marketing copy lose in both. Product page longevity SEO, in this environment, is no longer a question of how often the brand refreshes its copy. It is a question of how much of the customer's language the page is built to keep.

If you held to the standard model of a product page, the half-life of the page was about two years. After two years, organic traffic to the page was half what it was at peak, and the brand had to refresh the page to keep it ranking. This is the model under which most stores still operate.

Under the new model, the half-life of a product page is not a useful frame. The page does not decay. It accumulates. The page sent twenty visits a month in its first month, three hundred a month in its sixth, fifteen hundred a month in its twelfth, and is on track to send four thousand a month in its twenty-fourth, without the brand having refreshed the marketing copy once.

The brand that ships in this model is, in effect, paying a tiny operating cost to keep a system that reads, publishes, and replies, in exchange for an asset that no copywriter could have produced and no competitor can replicate without the same body of customers.

The brand that ships in the old model is, in effect, paying a copywriter to chase a sawtooth.

There is a quiet reading of this argument that the writer would like to leave the reader with.

A product page is, in its plainest description, a record of what a particular object means to the people who bought it. The marketing team's hero paragraph is a hypothesis about that meaning. The customers' reviews are the evidence. A page that is half hypothesis and half evidence, in 2026, is a page that wins the new search. A page that is all hypothesis, no matter how well-written, is a page that decays.

The half-life of a product page, in the model the engines now reward, is determined by whether the page learns. Pages that learn from their customers compound. Pages that do not, do not.

The work is to build the page that learns.