Is generative engine optimisation a distraction? If it means buying another dashboard, publishing an llms.txt file and rewriting useful pages so they sound unnatural to humans, probably yes.
But dismissing AI search altogether would be a mistake.
The current debate is often muddled because it treats two different questions as one:
- Does AI search send meaningful, attributable referral traffic today?
- Can appearing in AI answers influence how people discover, assess and remember a brand?
The answer to the first is largely no. The answer to the second is plausible, important and much harder to measure.
That means AI visibility should not replace SEO, conversion optimisation, digital PR or customer experience work. It should be an outcome of doing those jobs well.
The performance-channel case against GEO is strong
Marie Haynes’ recent analysis is the necessary dose of reality. Looking at 12 months of GA4 data across 100 sites with more than 145 million sessions, she found that ChatGPT traffic amounted to only 0.62% of Google organic traffic. Adding all tracked LLM referrals took the figure to just 0.73%. Her full analysis is worth reading. (mariehaynes.com)
The conversion data is more nuanced, but leads to the same practical conclusion. ChatGPT visitors converted at 13.8%, compared with 3.02% for Google organic visitors. That is a substantial difference in rate, yet scale matters: Google organic still delivered 96.5% of conversions in the dataset and generated 27 times more conversions than all LLMs combined. (mariehaynes.com)
There are important caveats. Only 55 of the 100 sites tracked conversions, while 45 did not track any key events at all. That is a measurement problem before it is an AI-search problem. But the available data is still enough to challenge a common sales pitch: most businesses cannot justify a major GEO budget on the basis of direct AI referral traffic.
Nor should they assume the line is heading inexorably upwards. Haynes found that collective ChatGPT referrals in her sample stopped their previous growth around the August 2025 release of GPT-5 and subsequently declined. (mariehaynes.com)
In short: do not take budget from proven organic search activity in pursuit of a speculative click stream.
The branding-channel case still deserves attention
Mike King makes the useful counterargument in AI Search Is Not a Distraction, It’s a Branding Channel. He agrees that the supposed GEO shortcuts — schema “hacks”, llms.txt files and content written chiefly for machines — are distractions. His point is that referral data is the wrong instrument for judging a channel where much of the activity happens before, or without, a click. (ipullrank.com)
King compares measuring AI search solely by referral visits with judging a Super Bowl advert by click-through rate. It is a deliberately provocative analogy, but the underlying point holds: people may ask an AI tool to compare providers, explain a category or narrow down options, then later search for a brand, visit directly or ask a colleague for a recommendation.
According to King, ChatGPT reaches 1.2 billion people weekly, Google’s AI Mode has passed one billion monthly users and AI Overviews reach 2.5 billion people. Those audience figures should be treated as platform-reported or industry estimates rather than proof of commercial impact, but they do make one thing clear: these are not fringe interfaces. (ipullrank.com)
The sensible position is therefore neither “GEO changes everything” nor “ignore AI until it sends clicks”. AI search is an emerging brand and consideration surface, while Google organic remains the core measurable acquisition channel.
That is consistent with the wider lesson of zero-click marketing: influence often happens on platforms that do not neatly hand the final click to your analytics package.
Stop optimising for robots and strengthen the evidence around your brand
The practical work is not exotic. It is the same steady work that supports organic visibility, conversion and reputation.
| Priority | What to do | Why it matters |
|---|---|---|
| Conversion pages | Build clear bottom-of-funnel landing and product pages | Makes any qualified visit more valuable |
| Independent validation | Earn coverage, links, reviews and inclusion in third-party lists | Gives buyers and systems evidence beyond your own claims |
| Brand consistency | Keep positioning, proof and product information accurate across channels | Reduces conflicting signals |
| Customer experience | Deliver support worth talking about | Reviews and discussion follow real experiences |
Build pages for buyers who are ready to act
If an AI answer does send someone to your website, they should not land on a vague thought-leadership article with no next step. Create landing pages for the commercial questions your buyers genuinely ask:
- What does the product or service solve?
- Who is it suitable for?
- What does implementation involve?
- How is it priced or scoped?
- What evidence supports the claims?
- What should a prospect do next?
A February 2026 PromptWatch analysis, shared in this post, examined nearly one million ChatGPT citations. Listicles represented 18.7% of cited content, with landing pages at 15.2% and product pages at 14%. Treat that as directional rather than a content formula, but it reinforces a useful point: commercial pages still need to be clear, indexable and genuinely useful. (linkedin.com)
A page built to convert is also a page that gives an AI system straightforward material to understand: what you offer, for whom, with what proof and under what conditions.
Aim to appear in other people’s comparisons
Publishing a self-promotional “best tools” list and putting yourself first has always been thin evidence. It looks even weaker when an AI system is trying to reconcile claims across multiple sources.
Dmytro Sokhach’s analysis of 11,025 commercial “best X” queries found that Google AI Overviews cited self-promotional listicles far less often in October 2026 than in July. When a brand did rank itself first in a cited listicle, it was also much less likely than a third-party number-one brand to appear among the answer’s top three recommendations. (aicitationinsights.com)
The takeaway is not “never publish comparison content”. Helpful comparisons can still serve visitors and support conventional SEO. The lesson is that you cannot publish your own praise and expect it to carry the authority of an independent recommendation.
Put effort into being included in credible editorial comparisons, trade publications, specialist newsletters, partner resources and community conversations where your buyers already look for advice.
Treat mentions and links as reputation infrastructure
Backlinks still matter for search, but their value is broader than a ranking signal. A relevant article that mentions your brand, explains what you do accurately and links to the right page can create referral traffic, familiarity and independent corroboration.
That is why digital PR and online authority should sit alongside content and technical SEO. The goal is not to manufacture citations in ChatGPT or Google AI Overviews. It is to create a trustworthy public record of what the business knows, delivers and stands for.
Useful inputs include original research, subject-matter expertise, customer evidence, well-supported opinions and practical tools. Empty “thought leadership” adds little to any channel.
Make customer support part of the visibility strategy
AI tools increasingly draw on the wider web, including review platforms, forums and other user-generated sources. The best way to improve that conversation is not to game it. It is to give customers fewer reasons to complain and more reasons to recommend you.
That requires prompt, capable support; honest expectation-setting; easy-to-understand policies; and a considered process for responding to criticism. Reviews are not merely a local SEO task delegated to the end of the quarter. They are public evidence of how the business behaves.
Support each commercial page with relevant social content
Social posts do not need to be a stream of generic brand updates. For each priority landing page, create useful platform-native material around the same buyer problem and language:
- a LinkedIn post explaining a common purchasing mistake;
- a short video answering an implementation concern;
- a customer story showing the outcome and limitations;
- a practical checklist that points back to the relevant service or product page.
The objective is consistency, not keyword stuffing. A buyer should encounter the same clear proposition whether they start with Google, ChatGPT, LinkedIn, a review site or an industry article.
Measure AI visibility without pretending it is paid search
Keep tracking LLM referrals, but do not mistake them for the whole story. At a minimum, track:
- AI referrals and their conversion quality;
- branded search demand;
- direct and homepage traffic trends;
- mentions, citation presence and sentiment for a defined set of buyer prompts;
- assisted conversions and self-reported “how did you hear about us?” responses.
Review these over quarters, not weekly dashboards. If possible, compare markets, product lines or campaign periods rather than assuming correlation proves causation.
The Edward Show discussion on the GEO myth, and its video version, reaches a similarly practical conclusion: much of what gets sold as a new discipline is simply sound SEO, content, brand and conversion work with fresh terminology.
The steady work wins
AI search is not a shortcut to visibility, nor a replacement for Google organic. Right now, the direct traffic evidence says it should not be funded as if it were a mature performance channel.
But brands should care about whether they appear accurately and credibly in AI-mediated research journeys. The investment is justified when it improves the foundations anyway: stronger landing pages, better proof, independent coverage, useful social content, customer satisfaction and reliable measurement.
There is no separate trick that makes a weak brand look trusted. The brands that compound visibility in AI search will be the ones doing the unglamorous work of becoming easier to understand, easier to verify and easier to recommend.
