Generative engine optimisation, or GEO, has produced a fast-growing measurement market filled with prompt tracking tools, share-of-voice dashboards, and AI visibility scores. None of those tools, on their own, can prove that a website change made a business better. They can sample whether a brand appeared in a set of AI answers, but they cannot connect a content edit to revenue. For ecommerce teams facing constant questions from leadership about AI readiness, the path forward runs through controlled testing, not more dashboards.
Why prompt tracking falls short as proof of impact
Prompt tracking platforms sample a slice of the prompts a user might type into ChatGPT, Perplexity, Gemini, AI Overviews, or AI Mode. They report brand mentions, citations, and visibility scores. That data is useful for debugging. It can flag early warning signs and help teams understand how a brand is described in certain contexts.
Prompt tracking tools are now expected to cover goals that belong to other functions. It is not the same as measuring the impact of a change on business outcomes. It is not the same as proving that a particular change should be rolled out across a large ecommerce catalogue.
Prompts are personal, often lengthy, and shaped by earlier exchanges. The same user may ask a follow-up that changes the context. Across a user base, answers can vary from one account to the next. The universe of possible prompts is so large that no known bound covers it. Every query differs, and no dashboard can record the full range of them. A sampled view cannot prove causation by itself.
Why ecommerce is a strong testing ground for GEO
AI assistants can research, compare, and recommend, but they cannot ship a product. The retailer, marketplace, or brand still plays the transactional role. That gives ecommerce teams a practical advantage: product and category pages remain essential, and they tend to live on scalable templates that can be tested.
Large ecommerce sites usually run on repeatable templates such as product detail pages (PDPs), product listing pages (PLPs), category pages, internal search pages, faceted pages, buying guides, and related content blocks. Those templates can be split into variant and control groups. Teams can change one element, compare performance, and measure what actually happened.
Online shopping is likely to grow in the coming years, and many purchase journeys will be shaped by organic discovery that blends search engine results with chatbot interactions. The interface may change, but the buying decision itself still ends up on the retailer’s site.
How LLMs actually find fresh product information
AI systems draw on two broad information sources. The first is training data, the language, entities, relationships, and brand context absorbed during model training. That knowledge is largely fixed until the next training run, which makes it a slow lever to pull.
The second source is retrieval, often described as retrieval-augmented generation, or RAG. When a user asks a question, the model may open dozens of background searches, read across the web, and synthesise an answer. For ecommerce, retrieval is essential. A model cannot rely on old training data to answer questions about current stock, today’s price, active discounts, latest reviews, delivery options, product availability, new launches, updated product details, or local availability.
Product feeds, structured data, PDPs, pricing, availability, and product attributes can matter for AI retrieval because they can help machines read a catalogue clearly. Anything that helps a machine read a catalogue clearly can also help AI assistants recommend it.
What fan-out queries change about optimisation
A user may type a single long prompt into an AI system, but the model may break that task into many background searches. It may search for product comparisons, reviews, pricing, availability, best options for a use case, brand reputation, delivery details, and other supporting information. The user sees one answer. Behind it, dozens of searches may have happened.
The old keyword-first model asked which keyword the team was targeting, where it ranked, and what the search result looked like. In AI discovery, hidden fan-out queries are where retrieval often happens. Teams usually cannot see all of the hidden fan-out queries, may get partial clues, and cannot treat the process as a clean keyword list.
Testing becomes more important for that reason. A team can make a change, then measure whether it improved LLM referrals, Google organic traffic, or the net business outcome. You do not need perfect visibility into every hidden query to measure the effect.
How to write a testable GEO hypothesis
Traditional SEO hypotheses usually work through one of three mechanisms: targeting new keywords, improving rankings for existing keywords, or changing the search result appearance to lift click-through rates. GEO has analogues, but the language shifts.
A GEO hypothesis might aim to target new fan-out queries, improve visibility for existing fan-out queries, influence the summary returned by an LLM, or make a page, product, or brand easier for the model to recommend. That fourth mechanism feels close to conversion rate optimisation, but the “converter” is partly the machine. The question becomes: has the page given the model enough information to confidently recommend the product?
That could mean clearer product suitability information, comparison language, reviews, freshness signals, structured data, key features, FAQs, delivery information, stock details, or buying guidance. A weak GEO hypothesis says: “This might help AI visibility.” A stronger one says: “Adding clearer product suitability information to PDPs may help models retrieve and recommend these products for more specific fan-out queries, while also improving confidence in the AI-generated summary.” That gives the team something measurable.
Where GEO testing happens on a large ecommerce site
GEO testing happens on the same scalable surfaces SEO testing already uses. That includes PDPs, PLPs, category templates, buying guide modules, comparison content, FAQs, review summaries, key feature summaries, internal linking modules, structured data freshness indicators, product feed-aligned content, and availability and delivery information.
The mechanics look familiar: a team makes a change to a variant group of pages, compares performance against a control group, and measures the result. What changes is the journey being measured. In traditional search, a user might open several tabs, compare sources, read reviews, check products, and then visit the site, with much of that research visible across a set of searches and visits. In AI discovery, more of that research happens inside the conversation. The model reads, compares, summarises, and narrows options before the user arrives, and the site may only see the final click. That can make the click more valuable, but also harder to interpret on its own.
Why GEO and SEO can disagree
Many GEO changes could plausibly help SEO. Changes such as more useful content, better structure, fresher product information, clearer summaries, stronger internal links, and better structured data can all carry SEO hypotheses. Overlap exists, but generative engine optimization and search engine optimization are not the same thing.
A change can help an LLM understand and summarise a page while reducing Google organic performance. A change that helps an LLM summarise a page can make the page bloated, duplicative, or less effective in traditional search. A team that looks only at LLM referrals and sees positive movement may find Google organic traffic drops by more than the LLM gain, leaving the business worse off.
What the SearchPilot and Omio test showed
The clearest example of GEO and SEO disagreeing comes from SearchPilot’s work with Omio, a travel comparison platform. SearchPilot partnered with Omio to share early GEO testing results with the wider industry, and the key lesson was that GEO and SEO do not always move together.
In one test, adding brand USPs increased LLM traffic by around 18 percent. In another, adding structured key takeaways performed positively for LLM-driven traffic, but would likely have hurt Google organic sessions by around 6.5 percent. Omio chose not to roll that change out and developed follow-up iterations instead. The AI result looked positive in isolation. The business result did not. Without measuring Google organic performance at the same time, the team could have shipped a net-negative change. That is the strongest argument against treating GEO as a checklist: a tactic can be directionally plausible and still wrong for a specific site, page type, or business goal.
What prompt tracking can actually tell teams
Prompt tracking keeps coming up as one of the most common tools ecommerce teams are adopting. Teams are not sure if they trust the data or what to do with it.
The better comparison is rank tracking in traditional SEO. Rank tracking is useful for debugging. It can help diagnose problems, show movement, surface early warning signs, and help teams understand where visibility may be changing. Prompt tracking has the same role, with extra complications. Teams do not know the full set of prompts users are typing. Many prompts are unique. Answers are personalised, and the model may use memory, previous conversations, location, and other context. A brand may appear for a prompt in one run and not another. A dashboard can only sample a small portion of reality.
Prompt tracking can help teams form hypotheses, flag issues, and account for some AI visibility patterns. It is not enough on its own to confirm that a GEO programme is working.
How to answer leadership on AI readiness
Executive attention on AI search has spiked. Leadership is asking whether the business is ready, what the team should prioritize, and how to accelerate progress. They deserve clearer answers than a dashboard of AI visibility scores.
The path forward is to frame GEO work as a series of testable hypotheses, run those tests on real page groups, measure AI and Google performance together, and report the net business impact. That gives leadership a concrete answer: here is what we changed, here is what we measured, and here is what rolled out or rolled back. Test results, not prompt screenshots, are what close the loop.
FAQ
Why is prompt tracking not enough to prove GEO is working?
Prompt tracking samples a small slice of an effectively infinite set of prompts, and answers are personalised, so it can show AI mentions but cannot connect a content change to business impact. Rank tracking in traditional SEO has the same limitation: useful for debugging, not for proving a rollout worked.
How can ecommerce teams test GEO changes?
Teams can run A/B tests on scalable templates such as PDPs, PLPs, category pages, buying guides, and FAQs by changing one element on a variant group, comparing against a control group, and measuring AI referrals, Google organic traffic, and revenue together.
Can a GEO change that helps LLMs hurt Google organic traffic?
Yes. SearchPilot’s work with Omio found that one change adding structured key takeaways performed positively for LLM-driven traffic but would likely have cut Google organic sessions by around 6.5 percent, so Omio did not roll it out. Measuring channels together is the only way to avoid shipping a net-negative change.
Want help with this in Myrtle Beach or Long Island? Talk to us.
NewSunSEO
Grow in Google and AI Search
AI-SEO, local search, paid ads and websites from one team, with offices on Long Island and in Myrtle Beach.
- AI-SEO
- Local AI-SEO
- AI Consulting
- PPC
- Web Design
- Web Development
- Long Island Office(631) 320-1700
- Myrtle Beach Office(843) 492-7131
Request a Free Proposal
Tell us about your business and where you want to grow.
We will reply by email to talk through your goals and what a proposal would include.
We use your details only to reply to you and send what you asked for. See our Privacy Policy.
