ع
A new company backed by Gulf investors is personalising merchants’ websites as AI changes how shoppers discover products, but measuring the result remains essential.
Keep Converting’s emergence from stealth in September offers a window into a changing part of online retail. As shoppers use AI services to research products, merchants are examining what happens after those services send a visitor to a store.
The company announced $2 million in pre-seed funding led by Nuwa Capital and COTU Ventures. Founded by Mohammad El Mougi and Manuel Prinz, it is building software that adapts the product-page experience to the visitor rather than presenting every shopper with an identical page.
The commercial question is larger than the funding round. If discovery increasingly happens outside a retailer’s website, the page receiving that visitor has a different job: it must connect a specific recommendation or question to the information needed to complete a purchase. More traffic alone does not establish that this handover works.
A new source of visitors
Adobe’s April 2026 analysis found that AI-referred traffic to US retail websites grew 393% year on year in the first quarter. In March, those visits converted into purchases at a rate 42% higher than non-AI traffic, reversing the relationship reported a year earlier. The analysis covers US retail, not a measured conversion rate for Gulf stores.
Adobe also examined whether website content could be read by AI systems. Product pages received an average readability score of 66%, below the 75% score for homepages. Those figures describe a separate problem from persuading a human visitor to buy: a product must first be visible and understandable to the system recommending it.
The two tasks can overlap, but they are not identical. A page may be easy for software to parse while leaving a shopper’s practical questions unanswered. It may also be visually persuasive while hiding important product information from automated systems. This creates two audiences for the same retail page, each encountering it in a different way.
What Keep Converting is trying to change
The company’s September announcement says its software generates different product-page experiences according to a visitor’s source and apparent intent. It reports deployments across the US, Europe and the Gulf, with integrations including Shopify, WooCommerce, Adobe Commerce, Zid and Salla.
Its founders bring experience from different parts of digital commerce. El Mougi previously worked at Vezeeta and Vodafone, while Prinz held product roles at TikTok Shop US and Walmart. The funding is intended to support merchant onboarding and engineering. These are the company’s disclosed plans and backgrounds, not proof that every merchant integration is active at scale.
The presence of Zid and Salla in the stated integration list gives the launch a direct connection to the Gulf’s merchant infrastructure. It does not reveal how many local retailers are paying customers or what share of revenue comes from the region. Those figures were not included in the launch announcement.
On its current website, Keep Converting describes variations in page layout, imagery and messaging, together with a process for merchant approval. It also lists subscription and performance-based commercial options. The distinction between generating a variant and approving its use matters when the page contains factual product information, delivery terms or other commitments made by the retailer.
The denominator behind the headline result
Keep Converting reports an average 64% increase in conversion rates across active client deployments. That is a company-reported result, not an independently verified forecast for a new customer. It also describes relative improvement, rather than a claim that 64% of all visitors buy something.
The arithmetic makes the distinction clear. A hypothetical store with a 2% conversion rate would reach 3.28% after a 64% relative increase, assuming the measures were otherwise comparable. That is an increase of 1.28 percentage points. It is an illustration of the calculation, not a reported result for any Keep Converting client.
Evaluating a combined average requires more information: the original conversion rates, number of participating stores, observation period and how individual results were weighted. The public headline does not establish whether the same improvement persisted across product categories or customer-acquisition channels. Nor does it, on its own, measure profit after fees and returns.
Those questions are normal to the economics of experimentation. A retailer can increase the proportion of visits resulting in an order while also changing order values or the mix of goods sold. Revenue per visitor and margin can therefore tell a different story from the conversion rate alone.
Automation does not remove the measurement problem
Shopify’s guide to multivariate testing, updated in August, explains that introducing more combinations increases the amount of traffic needed for a reliable result. Baseline conversion, the size of the improvement being tested and the distribution of traffic all affect the evidence required. A fixed number of days is not sufficient for every store.
Its guidance also recommends monitoring other outcomes alongside the principal measure, including average order value and returns. Those considerations remain relevant when software generates the alternatives automatically. Producing many versions quickly and establishing which version caused an improvement are different achievements.
For Gulf retailers, the immediate relevance of Keep Converting’s launch is the attempt to connect a changing discovery channel with existing merchant websites. The next useful evidence will be results broken out across stores and markets, with a clear account of what was measured. That would show whether the new product-page model changes commercial outcomes as well as the appearance of the page.
