Software brands publishing self-ranked comparison pages saw Google’s AI Overviews cite their content in 323 instances across 80 queries between April and June 2026, then recommend a competitor listed on the same page 69 percent of the time, according to research published by SEO consultant Lily Ray in June. The finding exposes a competitive liability in content strategies built around “best [category] software” listicles that rank the publishing brand first.
TL;DR: Lily Ray’s June 2026 study found that when Google AI Overviews cited a brand’s self-promotional listicle, the engine recommended a competitor from that same list in 224 out of 323 cases—leaving the publishing brand with the citation but no recommendation.
Ray analyzed 100 B2B software category queries through Google’s AI Overviews platform, testing each query three times between April and June 2026. Across the 80 queries that returned an AI Overview, the pattern held: brands earned citations for their own content but lost recommendations—and therefore customer action—to rivals mentioned within their own pages.
The research quantifies a structural problem in how generative search engines treat self-promotional content. Citation and recommendation represent two distinct outcomes in AI-generated answers, Ray’s data shows. A citation acknowledges a page as a source. A recommendation tells the user which product to choose. Only the latter drives conversions.
Citations Provide Visibility, Recommendations Drive Action
Google’s AI Overviews and similar platforms from ChatGPT and Perplexity separate informational attribution from transactional guidance. When a query triggers an AI Overview, the engine may cite five to eight sources in footnotes or inline references while recommending two to four specific products in the body of the answer.
Ray’s analysis found that software brands consistently appeared in citations—maintaining a presence on the screen—while recommendations went to competitors ranked within the same self-published listicle. For example, Oasis LMS appeared multiple times as a cited source for the query “best LMS for selling courses,” but Google recommended Kajabi, Thinkific, LearnWorlds, and Teachable instead. Each recommended brand was listed inside the Oasis article.
The distinction matters because users act on recommendations, not citations. A brand visible in the attribution stack without a recommendation in the answer text captures awareness but not intent. The recommendation is the extraction point where AI search engines convert query intent into a suggested action.
Third-Party Coverage Determines Which Brands Win Recommendations
Ray’s research identified the factor separating cited brands from recommended ones: external web coverage. Brands that won recommendations had significantly more referring domains and broader mention volume across AI platforms than brands cited but not recommended.
The gap does not stem from on-page optimization, the research shows. Self-promotional listicles trigger citations because they match query keywords and category terms. Recommendations, however, depend on what the broader web says about each brand—review volume on third-party sites, comparison articles from independent publishers, tutorial content on YouTube, and discussion threads on forums such as Reddit.
Google’s treatment of self-ranked pages has shifted, according to Ray’s findings. The engine now appears to weight external validation—independent reviews, user-generated comparisons, and third-party walkthroughs—more heavily than vendor claims when constructing the recommendation layer of an AI answer. Brands with limited coverage outside their own domain earn citations for their content but lose recommendations to competitors with larger footprints across external sources.
Measuring Citation Versus Recommendation Share of Voice
Ray outlined a five-step audit framework for tracking brand performance in AI search outputs. The process separates citation frequency from recommendation frequency, producing two metrics that standard rank-tracking tools do not distinguish.
Step one involves building a query list of buyer-intent phrases for the brand’s category: “best [category] software,” “[competitor name] alternatives,” and similar transactional searches. Step two requires running each query through Google’s AI Overviews and recording both the pages cited in attribution and the products recommended in the answer body. Step three calls for repeating each query multiple times, since AI-generated answers vary by session and location.
Step four calculates share of voice based on recommendation frequency rather than citation count. A brand mentioned in five out of ten AI answers but recommended in only one holds a 10 percent recommendation share, even if citation share reaches 50 percent. Step five extends the audit beyond Google to ChatGPT and Perplexity, mapping which publishers and content types each platform surfaces for category queries. The brand visibility framework published by AI search analytics providers in prior reporting follows similar logic.
Ray’s data showed consistency across categories. The pattern held for CRM software, help desk tools, learning management systems, and SEO platforms. In each vertical, self-promotional comparison pages generated citations while established brands with broader external coverage captured recommendations.
Third-Party Content Production as Competitive Response
The solution Ray’s research implies requires a structural shift in content strategy: increasing the volume of brand mentions on domains the vendor does not control. Reviews, comparisons, and walkthroughs published by third parties carry more weight in recommendation logic than equivalent content on the brand’s own site.
Traditional content marketing services focus output on owned properties—blog posts, landing pages, guides hosted on the company domain. AI recommendation dynamics favor unowned placements: a review on a software comparison site, a walkthrough on a creator’s YouTube channel, a thread on Reddit where users compare tools. The gap between citation and recommendation correlates with the gap between owned and earned content volume.
Ray’s findings suggest that affiliate programs—structured partnerships where third parties earn revenue for customer referrals—offer a mechanism for scaling external content production. Paying creators per conversion rather than per piece aligns incentive with output velocity. A revenue-share agreement produces ongoing coverage without per-article commissioning, the research notes.
The data does not specify which affiliate models perform best, but the underlying principle is clear: brands need consistent third-party mention velocity to compete for recommendations in AI search outputs. One-off placements generate isolated citations. Sustained external coverage shifts recommendation share.
Why This Matters Now
Australian software companies and service businesses investing in organic growth strategies face a content trap: the listicles and comparison pages that historically drove traffic now serve competitors in AI-generated answers. The June 2026 research from Ray quantifies what many businesses have observed anecdotally—content that ranks the publishing brand first earns the citation while sending the click to a rival.
The competitive implication is immediate. Self-promotional comparison pages remain cheap to produce and easy to scale, which explains their prevalence in B2B content strategies. But if 69 percent of AI Overview citations convert into competitor recommendations, the ROI calculation inverts. Brands pour resources into content that educates the market while a competitor with stronger external coverage captures the conversion. For businesses evaluating SEO consultation or content production investments, Ray’s data argues for shifting budget from owned comparison pages to earned external placements.
The tactical response involves measuring citation versus recommendation share now, before the gap widens further. Most Australian SMEs track whether their brand appears in AI answers; few separate informational attribution from transactional recommendation. The five-step audit Ray outlined costs nothing beyond time and produces two distinct metrics. Citation share reveals visibility. Recommendation share predicts revenue. The difference between them is the early warning.