Albert Zhu is the founder and CEO of Kongfuseo, working with e-commerce brands on search and AI visibility.
getty
Picture this: A shopper types “best sheepskin boots in Australia” into a generative AI tool and gets back four names, a price range and a note about sizing. None of the search results came from one product page. The tool pulled fragments of information from a review platform, a shopping guide and a magazine round-up, then wrote something that reads like advice from a friend who knows the category.
For a year, my strategy conversations with apparel clients have landed on the same question, usually about 20 minutes in: “Why does the AI tool recommend a competitor we outrank on Google?” And lately, they’ve been asking a harder one: “Why is our visibility rising on one AI platform while it falls on another?”
A Retailer’s Data, Pulled Apart
One of our clients, an Australian sheepskin footwear retailer with a largely domestic audience, demonstrates the point.
Using Ahrefs’ data on AI responses, I did some research. I found that as of September 2026, ChatGPT cited the client’s site in 869 responses across 858 distinct pages, up 645 responses and 756 pages month over month. Over the same 30 days, Google AI Mode citations fell by 385 to 678, AI Overviews citations fell by 106 to 569, and organic search traffic fell by 2,800 sessions to 28,200.
In my view, the direction matters more than the absolutes. The brand’s ChatGPT visibility roughly tripled while its Google AI visibility contracted and classic organic traffic declined—same site, same month, with no meaningful change to the link profile of backlinks from referring domains. Whatever moved the ChatGPT number, it was not links.
Before anyone considers building a plan on a single month, however, consider this caveat: Ahrefs has recently updated its organic traffic model, and its AI-response data is relatively new, so cross-period comparisons carry some methodological noise. The pattern has held across our other retail accounts, but one month is a signal, not proof.
The Concentration That More People Should Track
Divide distinct cited pages by total responses and you get a ratio worth watching. For this client’s site, ChatGPT sits at roughly 0.99: Almost every response reached for a different page. AI Mode is near 0.75, AI Overviews 0.49, and Gemini and Perplexity around 0.20 and 0.25, reusing a very small set of pages repeatedly.
That implies different work per platform. My hypothesis, based on observations about this client and others, is that winning on Perplexity or Gemini means getting a handful of pages cited over and over, and winning on ChatGPT means depth, because the model appears to keep reaching for a page it has not used yet.
I’ve noticed that many brands have the resources to optimize a small number of high-value pages. However, their AI visibility on some platforms may depend on having a broader body of useful content.
Mentions Tend To Outweigh Backlinks
The broader research points the same way. Ahrefs examined 75,000 brands with the Spearman correlation coefficient. Among its findings were that “brand web mentions show the strongest correlation (0.664) with AI Overview brand visibility” and “web mentions (0.664) correlate much more strongly than backlinks (0.218).” The next two strongest factors were also off-site: “branded anchors” and “branded search volume.”
Format matters, too. Ahrefs found that on ChatGPT, “across all source links, ‘best X’ blog lists represented 43.8% of all page types.” More specifically, Ahrefs discovered “a correlation between a brand positioning highly in third-party lists and being more likely to feature in response.”
Of course, correlation is not causation, and widely mentioned brands are usually brands that have spent a significant amount of time on their products and public relations. What the data suggests is that the evidence doesn’t live on your domain.
Where The Real Problem Sits
Analysis of my client’s data exposed something structural as well.
My analysis revealed that of the site’s 28,200 monthly organic visits, 25,500 come from branded queries and 2,200 from non-branded ones. The entity analysis is starker: The generic category term attracts 23,968 of the site’s keyword traffic, while the retailer’s own brand name accounts for 5,761, about 30%.
The category word does most of the work, and the category word belongs to someone else. In my experience, when a model resolves a query about that product type, it usually has abundant material about the category and little that fixes a particular retailer as a distinct entity. OpenAI’s documentation states that “merchants are ranked based on factors like availability, price, quality, and whether they are the maker or primary seller of that item.” That’s a difficult set of criteria for any multi-brand retailer selling what competitors also sell.
Best Practices For Retailers Optimizing For AI Recommendations
The fix is unglamorous. I recommend fixing one brand description and using it everywhere, such as a single sentence covering city, category, core materials, size range and returns policy, and then repeat it on the site, in press coverage, in reviews and in trade media. That way, when an AI solution or search engine encounters the company in different places, it can connect those mentions to the same brand identity, thanks to the recurring facts. Vague copy, such as a company having “a fresh, contemporary perspective,” isn’t really helpful.
Then make each placement distinct. Google’s documentation on AI features describes “a ‘query fan-out’ technique — issuing multiple related searches across subtopics and data sources — to develop a response.” Additionally, Google’s recent generative-search patent describes selecting “additional portions” of relevant source material based on “not being too similar to a portion already selected for the completeness set.”
Check your own house first, though. Eligibility for Google’s AI features starts with being indexed.
My takeaway is that every piece of earned coverage should carry a fact nobody else has, such as return rates and their causes, which sizes run true and delivery times by state. Journalists like getting exclusives. Based on my research, it appears that overall, so do generative AI models.
Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?
