The AI lie in most Amazon software tools - plus gaming ads as the new attention arbitrage
A veteran enterprise-tech CEO recently made a striking observation about the Amazon software ecosystem: about 99 percent of what is marketed as "AI" is actually just rule-based automation. Most "AI-powered" tools operate like this: if ACOS increases 1 percent, reduce the bid by 2 percent. That is not artificial intelligence - it is a conditional statement, an if-then rule that could be written in basic programming. Yet every agency and tool now claims AI-powered optimization, AI-driven insights, and AI everything.
The distinction matters because real AI learns patterns, adapts dynamically, and identifies insights humans miss. Consider the example making the rounds: when your competitor goes out of stock, your sales spike and your ACOS drops. A rule-based system interprets this as peak performance and doubles down. But you are not actually performing better - you are temporarily the only option available. True AI considers broader context: competitor inventory levels, market share fluctuations, seasonality patterns. It distinguishes genuine optimization from circumstantial advantage. That is the gap between automation and intelligence: one follows predetermined rules, the other analyzes and adapts. Caveat emptor the next time a tool sells you its AI.
The top question sellers are asking right now is how to extract all their reviews (or a competitor's) now that the major all-in-one seller tools have removed that option at Amazon's request. A workable path is a scraping marketplace - a platform where developers publish specialized scrapers, including dedicated Amazon review extractors. The workflow: search the marketplace for an Amazon reviews scraper with good ratings; configure it with your ASIN, the marketplace domain, and a sort order ("recent" for trends, "helpful" for top reviews); start with one page, about ten reviews, to test. Smart filters let you pull only 1-star reviews for complaints or 5-star for testimonials, search for terms like "durable" or "broke," or restrict to verified purchases.
Run the extraction (you can monitor progress and cost in real time and pause anytime), then export as JSON for AI tools or CSV/Excel for spreadsheets - each review includes title, text, rating, date, verified status, and media URLs. The payoff comes when you upload the export to an AI assistant with a prompt like: analyze these reviews and identify the 6 to 10 most frequently asked questions buyers have; for each, extract the concern, phrase it as a natural question, note frequency, and identify the buyer-journey stage, focusing on product fit, value, alternatives, compatibility, and ease of use. Keep costs down by starting small, using filters, scheduling regular small runs instead of one massive scrape, saving configurations, and double-checking ASINs before running. When Amazon blocks your review tools, specialized scrapers keep competitive research running, and AI turns the data into buyer intelligence.
On a related note, one seller recently demonstrated where Amazon keeps its richest data - the real reasons people return products (not the generic dropdown options), the searches that go nowhere because nobody has made the thing yet, the questions shoppers ask right before buying or bouncing, plus demographics and sentiment - and released a free way to capture it in one click, with AI that turns it into a playbook of product ideas, image strategy, and listings that speak to what people actually want. Worth tracking down in seller communities.
While most direct-to-consumer brands fight bidding wars for social placements, a massive shift is happening underneath: mobile gaming hit 47 billion downloads in 2025, and for Gen Z, gaming now beats TV and streaming in weekly hours (12.2 versus 11.8). Social media still commands 35.8 percent of mobile usage, but the quality of that attention is deteriorating. The real opportunity is not where people spend time - it is where they spend attention.
The psychology gap is the point. On social media, users are killing time: average engagement is 1.5 seconds before the scroll, the mental state is anxious and ad-blind, and your ad is an interruption they are trained to ignore. In gaming, users are in an active flow state with clear objectives - focused, achievement-oriented, engaged. A 15-second ad there is not something users tolerate; it is something they choose in exchange for power-ups, extra lives, or boosts. You are not interrupting their experience, you are enabling their win - and when you help someone achieve a goal, the psychological barrier to purchase drops.
The arbitrage window is open: while CPMs on Meta and TikTok keep climbing, mobile gaming offers one of the world's largest ad inventory pools that most e-commerce brands have not touched - massive supply, thin competition, users actually paying attention. Five tactics to win: link your call-to-action to the game reward ("Need a boost in real life? 20% off while you wait for your next life"); go longer, since 81 percent of gaming ad spend goes to ads of 31 seconds or more; hook in the first 5 seconds because many gaming ads are unskippable for about 7; rotate creative weekly with 3 to 5 variations because ad blindness hits faster in games; and test playable elements like swipe-to-reveal or choose-your-bundle, which hold attention longer than static video. Gaming ads are not a replacement for social - they are the arbitrage opportunity while everyone else fights for feed space, at pre-2018 social prices. The window will not stay open forever.
Quick hits: Meta's founder is teasing that agentic commerce is coming to its platforms; Walmart is allowing some sellers to incentivize reviews; TikTok is radically reshaping product discovery for the beauty industry; and Amazon sellers face new compliance challenges. Trivia answer: of the 13 billion packages Amazon delivered same- or next-day worldwide in 2025, 8 billion were in the USA.
