The 'listicle penalty' isn't about listicles

Sometime between Lily Ray's February 3, 2026 Substack post and the April–May trade press recaps, a narrative hardened: Google penalized listicles. Search Engine Land wrote it. Search Engine Roundtable wrote it. Sword and the Script wrote it. Half a dozen SEO newsletters wrote it. The phrase 'self-promotional listicle penalty' became the working shorthand for what the December 2025 broad core update did to a particular kind of B2B SaaS content.

The narrative is wrong. Or, more precisely, the narrative is a flattened version of an argument the person who started it has since publicly walked back. Ray's own March 18 follow-up was explicit: 'comparison content can absolutely be a legitimate and useful content format when done thoughtfully and in moderation. The problem isn't the tactic itself; it's the scale.' Her May 13 piece went further: 'the tools themselves are not the problem, but the implementation can be.' The originator of the data is not in the trade-press position the trade press attributes to her.

Devesh Khanal at Grow and Convert reached the same conclusion on May 26 with a piece titled 'Self-Promotional Listicles Aren't the Problem. Bad Content Is.' This essay starts from his thesis and pushes further, because the case has gotten more decisive than either Ray or Khanal has yet argued in public: there is a falsification test of the format hypothesis that nobody has applied, a technical counter-mechanism that nobody has engaged, and an existence proof for transparent self-promotion that has been hiding in plain sight for nine months. Together they make a stronger claim than 'listicles aren't the problem.' They make the claim: the signal Google is downweighting is unearned self-elevation, and the format is the scapegoat.

Ray's February 3 post is the canonical artifact in the listicle-penalty narrative. Reread it. She examined seven B2B SaaS sites that lost between 29% and 49% of SISTRIX U.S. visibility starting in mid-to-late January 2026 — measurement window after the December 2025 broad core update had completed. She did not name any of the seven (the largest was an '$8B B2B brand' with a −49% drop). She inventoried 191 self-promotional listicles across the affected subfolders. She ran originality.ai against samples and got 100% AI-generated confidence on multiple pages.

The signals she flagged were specific: company self-ranking #1 without independent verification; year-swap titles (2026 swapped in over a 2025 page with no substantive update); AI-generated text at scale; missing first-hand testing; programmatic templates; schema misuse, including fabricated AggregateRating; artificial dateModified refreshes. The pattern she described was a conjunction of behaviors, not a format.

Her March 18 piece corrected the trade-press distortion before it had fully crystallised: comparison content is legitimate; scale is the problem. Her May 13 piece corrected it again, explicitly: the tools are not the problem, the implementation is. Both posts were available before the April–May recaps that continued to cite her February 3 work as evidence of a categorical penalty. The originator of the data was already in a different place than the discourse built on top of it.

If you only read one Ray post, read May 13. If you've only read February 3, you have an out-of-date model of her position.

“The problem isn't the tactic itself; it's the scale.”

— Lily Ray · March 18, 2026

Seer Interactive published a 2 million-citation analysis on February 25, 2026 covering November 2025 through February 2026. They found two things at the same time. ChatGPT's listicle citations declined 30% month-over-month between December and January, dropping from 17.2% of citations to 15.5%. Google AI Overviews' listicle citations, measured against the same prompts, were 'trending up or staying flat.'

This is the cleanest possible falsification of the format hypothesis. If the signal Google and the LLMs were downweighting was the listicle as a structural form, both ChatGPT and AI Overviews would move in the same direction on identical inputs. They did not. Same listicle URLs. Same prompts. Opposite trajectories. Whatever changed in December and January, it cannot be the format itself — because the format is invariant across the two platforms and the platforms diverged.

Every published commentary on the Seer data treated the divergence as a curiosity ('listicles in AIOs are doing better, weird'). None of them treated it as what it actually is: a controlled experiment in which the format hypothesis fails. The hypothesis 'listicles are getting demoted because they're listicles' makes a prediction. The data refutes it. The remaining work is figuring out what the platforms were actually filtering on — and the answer has to be something downstream of the format itself.

On March 1, 2026, Carolyn Holzman published a forensic analysis arguing the demotion mechanism is technical, not editorial. Her case: Google's Helpful Content classifier treats jump-link fragment identifiers as duplicate-page signals, and on February 2, 2026 it demoted not just self-promotional B2B listicles but also Google's own developer documentation and Grokipedia — neither of which is a self-promotional anything. If a single classifier change hits a fragment-heavy listicle, Google's own developer docs, and a Wikipedia clone on the same day, the proximate cause is not 'AI is judging your editorial intent.' The proximate cause is a duplicate-detection rule firing on a shared structural feature.

She might be wrong. The mechanism might be partial, or one of several stacked changes. But the silence about her counter-mechanism is itself the finding. The SEO trade press reached for the narrative explanation — Google is cracking down on bad behavior — and skipped the technical one. The narrative explanation is more shareable; the technical one is more falsifiable. Independent writers should engage the falsifiable one even when it spoils the better story.

The two findings — Seer's divergence and Holzman's mechanism — together strip the case for a categorical listicle penalty down to almost nothing. The cross-platform behavior rules out a format-as-signal hypothesis. The Google-docs-and-Grokipedia comparison rules out a self-promotion-as-signal hypothesis at the level Holzman identifies. What's left for the format-penalty story to defend?

iPullRank publishes a piece called 'Top 11 Best Enterprise SEO Agencies of 2026.' iPullRank is listed at position #1. The first paragraph of the post reads: 'Obviously, we're going to kick off this listicle with ourselves. Yes, yes, we know, SEO faux pas.' The methodology section names the specific tool used to measure citations (Profound), the topic set, and the platforms covered. The author is Garrett Sussman; the framing is Mike King's. The page was published June 26, 2025 and last updated March 10, 2026 — that is, during the alleged crackdown, with the year-swap pattern the trade press flagged as a demotion signal.

The page is still live. It still ranks. Mike King is still publishing under his name. If 'self-promotional listicle ranked #1' were the signal Google's classifier downweighted, this page would have lost visibility on the exact dates Ray's anonymous casualties did. It didn't. The difference between iPullRank's #1 self-listing and the SaaS casualties Ray documented is not the format and not the self-placement. It is everything else: disclosed bias, named methodology, named author, real measurement, no template scaling.

Evertune published a 25,000-URL analysis on May 19, 2026 — after the listicle-penalty narrative had peaked. Of roughly 400 million citations across all models, 63% pointed to listicles. Half of the URLs Evertune analysed were listicles. Listicles, in May 2026, are still the dominant cited format in AI search. The narrative outran the data.

“Across nearly 400 million citations from all models, 63% pointed to listicles.”

— Evertune · May 19, 2026

Drop the format frame. The signal Google's classifier appears to act on is a conjunction — not any single property, but the co-occurrence of several. Read down the rows below and ask, against any given page, how many fire. Pages that fire none are fine. Pages that fire one or two are fine. Pages that fire five at once are the ones losing visibility.

These are the signals Lily Ray's casualties shared. They are the signals iPullRank's surviving page does not. They are the signals Google's May 15, 2026 Search Central guidance described as 'seeking inauthentic mentions across the web,' framing the problem as manipulation intent rather than form. They are also — and this is the part the trade press has not engaged with — the signals an honest publisher can choose to avoid without giving up the listicle format.

The reframe is not 'listicles are good.' It is 'these six signals are what's being penalized, and listicles are simply where they tend to cluster.' Avoid the signals; keep the format. The data says it works.