Eleven questions, real numbers, and the awkward advice that your best page may be the one where you admit a rival beats you.

1. Is the comparison page really the format AI assistants cite most?

Yes, with a caveat. Wix Studio’s AI Search Lab studied 75,000 AI answers and more than 1 million citations across ChatGPT, Google AI Mode and Perplexity (Search Engine Land, March 2026). Listicles took 21.9% of all citations, ahead of articles at 16.7% and product pages at 13.7%. On commercial questions, the “which one should I buy” kind, listicles took 40.9%. Glen Allsopp of Detailed, writing for Ahrefs, found “best X” lists were 43.8% of 26,283 ChatGPT source URLs. The caveat: Kevin Indig’s Growth Memo reports ChatGPT cut listicle citations by 30% from December 2025 to January 2026, and Peec AI notes newer ChatGPT retrieval leans toward product pages. The format leads. It is also being audited.

Table 1: Who Gets Cited, by Format

Study	Sample	What it found
Wix Studio AI Search Lab (Mar 2026)	75,000 answers, 1M+ citations	Listicles 21.9% of citations; 40.9% on commercial intent
Ahrefs / Glen Allsopp (Dec 2025)	26,283 ChatGPT source URLs	43.8% were “best X” lists
Scrunch (Sept 2026)	~8,000 classifiable AI-cited pages	33% were comparison or evaluation pages
Peec AI (Feb 2026)	1M+ citations	52% of cited listicles hit ChatGPT’s top citation tier

2. Why do assistants reach for this format?

Because a comparison page is pre-chewed. Ask an assistant “Asana or Monday for a 12-person agency?” and it has to produce a tradeoff. A page with named products, attribute rows and a verdict hands it one. BuzzStream’s study of 12,000 AI responses (June 2026) shows the pull: on head-to-head prompts, 35.8% of brand mentions came with a citation, against 7.2% on list prompts. Comparison questions are where assistants show their receipts.

3. Can I just rank myself #1 on my own “best of” list?

You can. It is the marketing version of voting for yourself for prom king, and the data says it buys less than you would hope.

Table 2: What Ranking Yourself First Actually Buys

Metric	Number	Source
Author brand recommended when its own listicle is cited	~4% rises to ~7%	Scrunch, ~10,000 URLs
Author brand mentioned (not recommended)	~19% rises to ~39%	Scrunch
Answers citing the listicle that recommend no brand at all	63.9%	Scrunch
A competitor recommended instead of the author	24.3%	Scrunch
Self-promotional share of listicle citations	~11% overall; ChatGPT 3.6%, AI Mode 10.3%, Perplexity 10.4%	Peec AI, 232,000 citations
Third-party vs. self-promotional lists (professional services)	80.9% vs. 19.1%	Wix Studio

Read that fourth row again. About one time in four, your self-flattering page did the research and the other guy got the recommendation.

4. Do all assistants treat these pages the same way?

No, and the spread is wide. Scrunch measured how often each surface recommends anyone after citing a neutral third-party comparison, and how often it recommends the author of a self-promoting one. Gemini hands out recommendations freely. Perplexity and Copilot are stingy with everybody.

Table 3: Recommendation Rates by AI Surface (Scrunch, May to June 2026)

AI surface	Neutral comparison cited: any brand recommended	Self-promoting listicle cited: author recommended
Gemini	64.1%	14.8%
Claude	49.9%	15.6%
ChatGPT	33.9%	8.6%
Google AI Overviews	28.7%	8.9%
Google AI Mode	25.8%	7.7%
Copilot	18.1%	4.8%
Perplexity	14.1%	6.3%

Across the board, the neutral page earns two to four times the recommendation rate of the self-serving one. Scrunch sells AI visibility monitoring and its causal sample is small, so treat these as directional.

5. What happened to the sites that leaned on it hardest?

Lily Ray tracked seven sites in SISTRIX data after the December 2025 core update. Visibility fell 29% to 49%. One $8 billion B2B brand with 191 self-promotional listicles lost 49% between January 21 and February 2, 2026. Her summary: “it works, until it doesn’t.” On May 15, 2026, Google confirmed its spam policies apply “to all of Google Search, including generative AI responses.” Growth Memo’s 5.32 million search results add nuance: publishers beat vendor listicles in 54.0% of 4,026 direct matchups, yet 62% of vendor listicles still gained traffic since January, with a median gain of 38%. Good ones survive. Lazy ones get found out.

6. Why on earth would I name where I lose?

1. Buyers can’t tell you apart. Wynter surveyed 100 B2B SaaS marketing leaders in June 2026. 98% said competing vendors at least sometimes seem to do the same thing, 64% found them hard to distinguish from their websites, and in a blind value-proposition matching test they scored 1.86 out of 5. Random guessing scores 1.0.

2. Perfect looks fake. Northwestern’s Spiegel Research Center and PowerReviews found purchase likelihood peaks at a 4.2 to 4.5 star average, then drops as ratings approach 5.0.

3. Competition helps when it shows up early. Gong Labs found deals where competitors surface early have a 49% greater chance of closing than deals with no competitor at all.

4. It is positioning. April Dunford starts positioning with “competitive alternatives,” meaning what customers would use if you did not exist. Saying who you are wrong for tells the right buyer, and the assistant, who you are right for.

Table 4: An Illustrative “Where We Lose” Block (Hypothetical Brands)

	Acme CRM (us)	BigCo CRM (them)
Best for	Teams of 5 to 50 that want setup in a day	Enterprises with a dedicated admin
Where we win	Price ($29 vs. $75 per seat), onboarding time	Custom reporting, 400+ integrations
Where we lose	No offline mode; reporting is basic	Takes weeks to configure

7. What goes into a comparison page worth citing?

1. Real competitors, by name. Scrunch found pages listing 4 to 10 brands earned the author more recommendations than pages with 2 to 3 or 11 to 15.

2. An honest ranking. Ahrefs noted DoorLoop placed itself 7th on its own list. That reads as a judgment call instead of a brochure.

3. A “choose them if” line for every rival, plus a table with the same rows for everyone.

4. Dated, sourced facts. Ahrefs found 79.1% of cited lists had been updated in 2025 and 26% within the prior two months.

5. Stated method and disclosure. Say you make one of the products. Readers know anyway.

The payoff is commercial too. Grow and Convert reports comparison and alternative keywords converting at 8.43% on average, versus 4.85% for category keywords and under half a percent for typical SEO content.

8. Most citations come from other sites. Does my own page still matter?

It does, as the fact sheet everyone else copies. Aleyda Solis (SEOFOMO) analyzed 15 SaaS brands in July 2026 and found 84% to 93% of AI citation weight comes from outside a brand’s own site. Comparisons and listicles made up only 2.6% to 5.2% of owned-page citation weight in ChatGPT and 7.2% to 7.6% in AI Mode. Meanwhile Peec AI’s study of nearly 200,000 responses found that sitting at #1 on a frequently cited third-party listicle goes with a 16.5 point higher mention probability in B2B SaaS. So publish the honest page, then earn spots on lists you do not own.

9. What are the legal limits?

I am a marketer, so treat this as a map and have counsel review anything aggressive.

Table 5: The Legal Guardrails

Rule	What it says	Practical translation
FTC, 16 CFR 14.15 (US)	Comparative ads are encouraged; disparagement is fine “so long as it is truthful and not deceptive”	Name names. Keep proof on file for every claim.
Lanham Act §43(a) (US)	Competitors can sue over false or misleading claims; liability is strict, and upheld jury awards “regularly reach tens of millions”	A stale pricing row can be “literally false.” Intent is no defense.
EU Directive 2006/114/EC, Art. 4	Compare “material, relevant, verifiable and representative features”; do not denigrate	Stick to checkable facts. Skip the insults.
Google spam policies (May 2026)	Now cover AI Overviews and AI Mode	Mass-produced “best” pages are a liability.

One cautionary tale from the New York State Bar Association: Kangadis Food was hit over a “100% Pure Olive Oil” label, settled, and filed for Chapter 11 about a year later citing legal fees. Use rival names in plain text and leave their logos alone.

10. And the practical limits?

Comparison pages rot. Competitors change pricing quarterly, so put an “as of” date on every table and screenshot your sources. You also cannot fake independence on your own domain, and assistants weigh the page accordingly. The IAB reports only 16% of brands track AI visibility systematically (per PPC Land), which means most teams publishing these pages have no idea whether they are being recommended or merely mentioned.

11. What should I do next week?

Pick your top three rivals. Write one page per matchup with a “where we lose” block, a date and sources. Then ask five assistants the head-to-head question and record who gets recommended, which is the number that pays the bills.

Sources (* = from the 1000startups.com independent source list)

1. * Peec AI, content types cited most by engine

2. * Peec AI, self-promotional listicles (232k citations)

3. * Peec AI, the listicle rank effect

4. * Growth Memo (Kevin Indig), Has Google demoted listicles?

5. * Detailed (Glen Allsopp) for Ahrefs, “best” lists research

6. * Wynter, B2B vendor differentiation survey

7. * Grow and Convert, bottom-of-the-funnel keywords

8. * Gong Labs, winning competitive deals

9. * SEOFOMO (Aleyda Solis), SaaS AI search findings

10. * April Dunford, positioning and competitive alternatives

11. * SISTRIX, Visibility Index data (via Lily Ray)

12. Search Engine Land, Wix Studio AI Search Lab study

13. PPC Land, Scrunch self-ranked brands study

14. Lily Ray, Is Google cracking down on self-promotional listicles?

15. Search Engine Land, Growth Memo listicle study coverage

16. BuzzStream, AI mentions vs. citations study

17. PowerReviews and Northwestern Spiegel Research Center

18. FTC, 16 CFR 14.15, comparative advertising

19. NYSBA, the Lanham Act and online advertising

20. EU Directive 2006/114/EC

21. PPC Land, Google spam policies cover AI Overviews and AI Mode