Ask three different systems what kind of page deserves to win right now, and in August 2026 you get three different, well-documented answers, all measured independently within the same few weeks. Google's classic rankings are pulling back from listicles. ChatGPT is still citing them at nearly the rate it always has, after already cutting back once. And Google's own AI Mode is quietly reallocating its citation pool away from the publishers who write listicles and toward the retailers whose products get listed in them. None of these three findings are new. Read together, for the first time, they are.
This piece synthesizes three separately published data sets from August 2026 into one picture: what happens to a content strategy built on comparison and best-of formats when the systems judging it are moving in three different directions simultaneously, and what a team should actually build in response, rather than which single metric to watch more closely.
Methodology
The organic-ranking findings come from a Growth Memo study published August 24, 2026, analyzing 60,000 US-English desktop queries across 15 verticals -- 5.32 million organic-result rows in total, gathered in a 47-hour window in August 2026. A secondary comparison held 2,400 fixed queries constant between January and August 2026, covering 24,000 search results, to isolate month-over-month movement rather than query-set drift. The AI Mode reallocation findings come from Growth Memo's Growth Intelligence Brief #23, published August 6, 2026, drawing on the same firm's Search Signals Index, which tracks roughly 2,600 companies across 26 verticals, measuring both SEO visibility and AI mention volume, with the AI Mode window covering June 15 to July 15, 2026. Something Inc.'s own citation-format data, cited here as previously published, comes from our tracking of a few thousand high-intent B2B queries across ChatGPT, Perplexity, Claude, and Google AI Mode. Every figure below is attributed to its source at first use; nothing here is presented as a single unified study, because it isn't one.
Google is quietly demoting listicles
Listicles are not disappearing from Google. They're still everywhere: 55.1% of the search-results pages in the Growth Memo sample contained at least one true numbered listicle, and 85.9% contained a page with a broader list cue -- a number or a word like "best" or "top" in the title. When a query explicitly asks for a set of options, Google shows 4.5 times more listicles than it does for other query types. This is not a format in decline by volume.
It is, however, a format losing ground at the positions that matter most. Comparing January 2026 to August 2026 on the same 2,400 fixed queries, listicles' share of the position 1 slot fell from a range of 18.9-20.2% down to 15.1-15.6% -- a drop of roughly four percentage points, or something like a fifth of their prior share at the top spot. The top-3 share moved similarly, from 35.9-39.0% down to 32.0-33.0%. Put plainly: almost 4% fewer listicles are winning position 1 than were eight months earlier, on the exact same query set.
Listicle share of Google's top organic positions, Jan 2026 vs Aug 2026, 2,400 fixed queries (Growth Memo)
The study also isolated a head-to-head effect worth naming precisely: when a publisher's listicle and a brand's self-serving listicle compete directly for the same query, the publisher wins 54% of those matchups, with a stronger average conditional rank (3.76 versus 4.65 for the brand page, on a 10-point scale). Google is not simply penalizing the listicle format. It looks closer to a trust adjustment against self-interested list-making specifically, layered on top of a smaller, broader pullback from the format overall.
| VERTICAL | LISTICLES IN TOP 10 | SELF-SERVING BRAND LISTICLES |
|---|---|---|
| B2B software | 62.5% | 46.2% |
| Beauty / fashion | 67.2% | not isolated |
| Entertainment / gaming | 42.0% | not isolated |
| Food | not isolated | 3.3% |
The vertical spread matters for anyone deciding where to spend on this format. B2B software carries both the highest listicle density in the top 10 and one of the highest shares of self-serving brand listicles competing in that mix -- meaning it's also the vertical most exposed to the trust penalty against brand-authored comparisons. Food sits at the opposite end, where branded listicles barely register. And across the sample, 92% of listicle-heavy SERPs also included a Reddit or YouTube result, which lines up with the six link types AI engines actually trust: community and video content keep showing up as load-bearing corroboration, not just as competition.
One more piece of context matters here: AI Overviews showed up in 83.7% to 93.4% of the results pages in the same sample, peaking in the B2B vertical specifically. That means the vertical under the most organic pressure against its listicles -- B2B software -- is also the vertical where an AI-generated summary is most likely to sit above whatever survives the organic reshuffle. Losing position 1 in that environment costs more than it used to, because fewer people are scrolling past the AI Overview to see who holds it.
AI Mode is quietly reallocating who it cites
The second data set is a different kind of shift: not which format wins, but which kind of entity gets named at all. Growth Intelligence Brief #23 found that across a 30-day window from June 15 to July 15, 2026, Google AI Mode's total mention volume contracted 12.9% across its tracked index, ending the period at roughly 626,000 mentions per week. A shrinking pool is one story. What happened inside that shrinking pool is the more important one: publisher mentions were cut nearly in half over the same window, while named retailers moved sharply the other way. Home Depot's mentions roughly tripled. Wayfair's roughly quadrupled. And the shift from informational to commercial framing wasn't isolated to home goods -- it ran across 15 of the 20 verticals the index tracks.
The timing lines up with the tail end of Google's June 2026 spam update, which the brief notes completed on June 26 -- squarely inside the measurement window. That doesn't prove causation on its own, but it's a reasonable place to start looking if you're trying to explain why an update aimed at classic search spam appears to have reshaped which kinds of entities an AI answer surface considers worth naming.
It's also worth sitting with what a 15-of-20-verticals spread actually implies. This isn't a home-goods-specific story that happens to feature Home Depot and Wayfair as convenient examples. Three-quarters of the tracked verticals moved the same direction over the same window, which is the signature of a systemic scoring change rather than a handful of individual brands having a strong month. Whatever shifted inside AI Mode's retrieval or ranking logic in June, it wasn't tuned for furniture and hardware. Those two brands are simply where the shift happened to produce the largest, easiest-to-name swings.
“A shrinking citation pool that's also reallocating toward commercial entities is two bad trends for publishers happening at once, not one.”
Where the two trends collide
Here's the part that doesn't show up in either study on its own. We've previously found, across our own tracking, that comparison and alternatives content earns roughly a third of all AI citations -- more than any other single format, because it answers the buyer's first real question directly. That finding still holds. Separately, ChatGPT's own citation mix, per third-party tracking, still runs at 43.8% listicle-format pages, drawn from a sample of over 26,000 source URLs -- even after ChatGPT itself cut its listicle-citation rate by roughly 30% between December 2025 and January 2026. The format took a hit and is still, by a wide margin, the single largest category of what gets cited.
Put the three data points next to each other and the picture sharpens. List and comparison content is still the best-performing format for earning an AI citation, full stop. But the specific pages winning that format are under pressure from two directions at once: Google's organic rankings are pulling back from listicles generally and penalizing self-serving ones specifically, while Google AI Mode is separately reallocating its citation pool away from the publisher-style pages that have historically produced the highest-performing listicles, toward commercial entities directly. A publisher's best-of page, historically the strongest format for both organic ranking and AI citation, is now exposed on both fronts at once: losing organic position to a stricter trust bar, and losing AI Mode citation share simply for being a publisher rather than a named commercial entity.
This is also the throughline connecting our own prior work on content chunking and length by engine and on why refreshed content keeps winning citations over merely old content: format alone was never going to be a durable answer, because the systems scoring format keep moving independently of each other. The August 2026 data adds a new axis to that instability -- not just format and freshness, but the entity type behind the page.
It's worth being precise about what these three data sets do and don't share. They weren't run by the same team, on the same query set, or with the same stated goal -- one measures organic position, one measures AI Mode's internal citation pool, and one measures cross-engine citation share from our own tracking. None of them set out to describe the other two. That's exactly why the overlap is worth taking seriously rather than dismissing as coincidence: three independent measurement efforts, run by different organizations with different incentives, converged on the same underlying story in the same month without coordinating on it.
There's also a scale mismatch worth naming plainly. A 4.4-point swing in listicle position-1 share is a meaningful but gradual move -- the kind of change a quarterly audit would eventually catch. A near-halving of publisher mentions inside a single 30-day AI Mode window is not gradual. If your reporting cadence is built around the slower signal, you will miss the faster one until a client asks why AI-cited traffic to your best-of pages fell off in a month you weren't specifically watching.
What to build now
The intersection of all three data sets points to a specific kind of page, not a specific format in the abstract. Comparison and listicle-structured content still wins the most AI citations of any format, so don't abandon it. But build it from a position that survives both pressures identified here: attach it to a real commercial or product entity rather than pure informational authority, since that's the direction AI Mode's citation pool is moving, and hold it to the same evidentiary bar the trust penalty implies -- sourced claims, a genuine head-to-head, not a self-interested ranking dressed as objective comparison.
For ecommerce and retail brands specifically, this is closer to good news than bad: you're the entity type gaining AI Mode mention share, and a well-built comparison page under your own domain captures both the organic listicle demand that still exists and the AI citation demand that still favors the format, without inheriting the trust penalty aimed at brand-authored best-of pages, because you're comparing your own product line honestly rather than ranking competitors. It's the same instinct behind how Johann Wolff built 40-plus frame and style guides into a durable organic and citation asset: comparison and guide content anchored to real product depth, not aggregation.
For publishers and informational sites, the read is more defensive. The format that used to be your best citation asset is now competing against retailers for the exact same citation slots, in a shrinking pool. The response isn't to abandon comparison content -- it's still the highest-leverage format available -- but to build it with the same evidentiary rigor a hub-and-spoke content architecture demands: named sources, real testing or data behind every ranking claim, and enough independent authority that a trust-sensitive algorithm doesn't lump you in with self-serving brand lists. Under content marketing built for both surfaces and generative engine optimization together, that's a two-signal build, not a one-time content sprint: rank for the comparison query and survive the citation reallocation happening underneath it.
Three concrete moves follow from the data rather than from general GEO advice. First, audit which of your existing comparison and best-of pages are brand-authored versus independently reported, because the 54% publisher-win-rate finding suggests the independent framing itself is worth something structurally, not just as a trust signal to readers. Second, if you're a retailer or product brand, don't wait for a publisher to write the comparison that features you -- build your own, honestly, because the AI Mode reallocation data says the citation slot is moving toward entities like you regardless of who writes the best version of it. Third, treat the two timelines separately in your reporting: organic listicle share on a quarterly cadence, AI Mode mention composition on a monthly one, because collapsing them into a single dashboard number would hide exactly the divergence this research describes.
None of this is a call to chase Google's or OpenAI's internal logic query by query. It's a call to notice when three independent measurement efforts agree that the ground moved, and to check your own content against all three axes -- format, freshness, and entity type -- rather than the one axis you happened to be tracking already. A comparison page built and measured on format alone was a reasonable bet as recently as this spring. By this data, it no longer is.
Cite this research
Source data: Growth Memo, "Has Google demoted listicles?", August 24, 2026; Growth Memo, "Growth Intelligence Brief #23", August 6, 2026. Reuse the framing freely, with attribution back to the original studies and to this synthesis.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Tyler leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.