Every week we get asked the same question by Amazon sellers: can AI replace what an agency does? Can a few clever prompts to Claude or ChatGPT take the place of years of operating experience and proper enterprise tools?
The honest answer is no, but it is also not a simple no. AI has become a serious part of how we run Ecommerce Intelligence and a second Anthropic-built business behind the scenes. It has changed how we build, how we problem-solve, and how we deliver value to customers. What it has not done is replace the data, judgement, and execution that actually grow Amazon brands.
This article breaks down where AI genuinely falls short for Amazon sellers, and where it delivers real value when used correctly.
Why AI Cannot Replace Amazon Operating Experience
There is a growing belief that LLMs can run an Amazon strategy end-to-end. After two years of using Claude and ChatGPT daily across two operating businesses, the limits are very clear. Below are the areas where AI consistently underperforms.
LLMs Are Blocked From Crawling Amazon
This is the single biggest limitation most sellers do not realise. LLMs are currently blocked from crawling Amazon, which means no data-led insights can be pulled directly from the platform. You have to feed the model the right data at the right time, and knowing what “right” looks like is itself the expertise.
This is why agency experience and enterprise tooling still matter. Without the right data going in, you only get generic frameworks coming out.
Framework Advice Is Not the Same as Strategy
Claude and ChatGPT give solid framework-level advertising advice. Ask either of them how to structure a Sponsored Products campaign and you will get a reasonable answer. The problem is that frameworks alone are not strategy.
Without enterprise tools plugged into the Amazon Ads API, operated by experienced teams who know which data to extract, your output will fall behind competitors who invest in both infrastructure and people. Prompts alone will not keep pace with brands running proper data pipelines.
Poor Advice Circulates Widely
One example we have seen repeatedly is the recommendation to “price your product within 5% of the lowest competitor.” The source was a random US agency, and the advice has been picked up and repeated by LLMs as if it were settled best practice.
It is horrendous advice because it ignores your own margins, your data, and your product performance entirely. This is the danger of treating AI output as authoritative without questioning where it came from.
Hallucinations and Legacy Information Still Happen
All LLMs still hallucinate or surface legacy information, some of it up to five years old. This still happens with Claude Opus released on April 26, which is the most capable model currently available.
The same applies to navigation paths. Ask an LLM how to find a setting in Seller Central and it will often confidently tell you “dashboard > inventory > add product” based on outdated UI. In complex API and developer systems, this has cost real time and caused real headaches.
Listing Creative Without Performance Monitoring Is Pointless
LLMs can produce excellent listing creative and beautiful A+ and storefront pages. The output is genuinely good. The problem is that performance has to be analysed rather than just admired.
Without active CTR and CVR monitoring with human adjustments to move the needle, even the best creative output leaves you obsolete against larger competitors who run ongoing testing. Obvious AI imagery can also hurt conversion rate, so visuals must look real and genuinely improve on existing photography rather than replace it. Listing optimisation is a long-term refinement process over months, not a one-off creative output, and executing AI-led work often takes longer than investing the same time in strategic activity that already works.
Catalogue and Brand Issues Are Beyond AI’s Reach
Catalogue and brand issues are central to Amazon success. AI can give loose guidance where information is publicly documented, but diagnosis through these tools typically takes longer than the actual fix.
This is especially true when Seller Support does not understand the issue and it needs higher escalation. An LLM cannot escalate a case, push back on a generic response, or know when to stop using the standard support route entirely.
Default Sources Are US-Centric
Claude’s default sources are US-centric, so recommendations are not always valid in the UK and EU. We have seen it recommend MAP (Minimum Advertised Price) policies and other practices that are illegal in this market. If you are selling in the UK or EU and acting on AI advice without verifying it locally, you can end up exposed.
Where AI Actually Delivers Value for Amazon Brands
The picture is not one-sided. AI has become genuinely useful in our day-to-day operations once you understand where to apply it.
Building Tools, Models and Web Platforms
The biggest strength of AI is its ability to simplify complex processes by building tools, models and web platforms. Ecommerce Intelligence has invested heavily in this area, and it is the part of the technology that has changed how the business runs.
It creates more areas of value for customers and frees the team from laborious, process-led tasks. Tasks that used to take hours of manual work now run inside internal tools we have built ourselves.
Overcoming Bottlenecks and Generating Ideas
When you are stuck working through a problem, AI is excellent for generating solutions or fresh ideas. This is the use case where Claude is our recommendation.
Whether you are trying to interpret a piece of data, work out a new campaign structure, or unblock a technical issue, having a fast thinking partner makes a real difference. It does not replace the strategist, but it speeds up the strategist’s thinking.
Maximum-Character Keyword-Driven Listing Creative
LLMs can produce maximum-character, keyword-driven listing creative on par with what competitors are putting out. This is the area where the gap between AI and human writers has closed the most.
The caveat from earlier still applies. Strong copy without performance monitoring will not save you, but as a starting point for new listings or a refresh, the output is genuinely competitive.
Lifelike Image and Video Generation
Image and video generation is now genuinely lifelike. The technology has moved a long way in twelve months.
The catch is that every frame and still has to be checked to ensure the product is shown accurately. Hands, packaging, branding, text on labels, and proportions are all areas where mistakes still slip through, and on Amazon those mistakes will hurt conversion.
Producing Whatever You Need, As Long As You Bring the Plan
AI can produce essentially anything you ask of it, provided you bring the plan and can accurately interpret the data you feed it. This is the core principle that separates sellers who get value from AI from those who do not.
Without a plan, you get generic output. With a plan and accurate data, you get a real productivity multiplier.
Processing Large Text Datasets and Contracts
AI is exceptional at processing large text datasets, interpreting contracts, and defining key clauses or heads of terms. This has become one of the highest-value uses of the technology in our operations.
Reviewing supplier agreements, parsing through customer feedback at scale, or summarising long-form research is where LLMs save the most hours. It is also low-risk because the source material is always in your hands to verify against.
The Key Takeaway
AI is a serious tool when it is treated as a tool, not a strategy. The sellers who get the most value from it are the ones who already know what good looks like on Amazon, who have the data infrastructure to feed it properly, and who use it to remove bottlenecks rather than replace judgement.
The sellers who get hurt are the ones who treat AI as an authoritative voice on a platform it cannot even crawl. Frameworks are not strategy, hallucinated advice can damage a brand, and listing creative without performance monitoring is just words on a page.
If you want to understand which growth levers actually work in 2026, our breakdown of Amazon growth levers is a good place to start. And if you would like a proper review of where AI fits into your operation rather than where it does not, get in touch with the team.
