AI writing assistants: what AI recommends, November 2026

2Where the AIs disagree

Among the top 15, the brands with the largest gap between the AI that recommended them most and the AI that recommended them least.

Do the AIs recommend themselves?

Some of the products recommended here are made by the AIs answering. Each row is one AI recommending one AI maker’s products: in how many of its decision answers, how often it had searched the web, and how often the maker’s own site was among its sources. Rows where an AI recommends its own maker come first.

When ChatGPT recommended itself (13 of its 46 decision answers), the pages its own search found named it every time (12 by the page's text, 1 by its address or title only); 76.9% of those answers had a page on OpenAI’s own site among the sources.

When Claude recommended itself (24 of its 45 decision answers), the pages its own search found named it in 95.8% of those answers (22 by the page's text, 1 by its address or title only); 4.2% of those answers had a page on Anthropic’s own site among the sources.

When Gemini recommended itself (3 of its 46 decision answers), the pages its own search found never named it.

When Perplexity recommended itself (1 of its 46 decision answers), the pages its own search found named it every time (1 by its address or title only); 100% of those answers had a page on Perplexity’s own site among the sources.

ChatGPT was recommended in 39.0% of decision answers by all four AIs, and in 42.5% by the other three; Claude was recommended in 35.1% of decision answers by all four AIs, and in 29.1% by the other three; Gemini was recommended in 4.9% of decision answers by all four AIs, and in 4.4% by the other three.

3How each AI answers

Gemini answered 17 of 60 questions without running any web search. That is the model Citewake asked, gemini-3.1-flash-lite, through its API with search available. Across all four AIs, 640 searches were run; 32.2% of them included a year. Claude answered 1 question without searching. Every page Gemini reported was one it cited, so its pages found and cited match.

5Agreement

On 42 of the 60 questions all four AIs recommended a brand first; they named the same brand first on 7.

6Method in brief

• Questions. 60 scored questions in three tiers (head, mid, tail), written as a buyer would ask them and never naming a brand: 46 decision questions and 13 informational. See the list. A further 15 validation questions are kept private.
• Asking. Each question was asked once of each AI through its official API, with its own web search on and no location sent. Models: ChatGPT: gpt-6-luna; Claude: claude-haiku-5-5; Gemini: gemini-3.1-flash-lite; Perplexity: perplexity/sonar.
• Collection. 29 October 2026, 09:19–09:38 UTC. 239 valid answers of 240 planned: 203 named at least one brand, 36 named none, 1 were refused or off topic.
• Reading. One grand master model (Claude Opus 5.5) read the four answers to each question side by side and marked each brand as recommended, mentioned or negative. A brand counts only if its name appears word for word in the answer.
• Measure. Recommendation rate is the share of decision answers in which the AI recommended the brand. The overall rate weights each AI and each tier equally, so it can differ slightly from a raw count. Intervals are 90%, from a cluster bootstrap that resamples whole questions. Here the base is 183 decision answers (46 decision questions).
• Method. v0.8, benchmark v1, scoring run 20261029T102417193Z. The full method: citewake.com/methodology.

7Limitations

• One run. Each question was asked once of each AI, in one 19-minute window. AI answers vary from run to run; the intervals cover variation across questions, not repeat runs.
• English, no market. Questions were in English and no location was sent, so the results describe no particular country.
• Spellings merged. claudeai was counted as Claude; jasperai was counted as Jasper; jenniai was counted as Jenni AI; perplexityai was counted as Perplexity; surferai was counted as Surfer; copyai was counted as Copy.ai; eeselai was counted as eesel AI; sciteai was counted as Scite; riversidefm was counted as Riverside; lavenderai was counted as lavender; trinkaai was counted as trinka.
• Benchmark figures, not market share. These are benchmark visibility figures across a fixed bank of questions with equal weight per tier, not estimates of market share.
• Self-recommendation sources. Whether the pages an AI found named a product is read from their fetched text where Citewake has it, and otherwise from the page’s address and title alone (the two are counted apart); where none of an AI’s pages can be read, it is shown as not known, never as zero. Decision answers only; cross-category questions are left out.
• The AIs in the ranking. ChatGPT, Claude, Gemini and Perplexity also answer the questions; where one of them is recommended, its rate from the other three AIs is shown beside it.
• Top picks are rare. Most answers do not name a single best choice. Even ChatGPT was the top pick in only 8.8% of decision answers, so this report ranks by recommendation rate.

8Disclosure

Citewake is operated by the team that also builds aipozi, Biotree and livedot. Biotree and livedot appear in these categories. They were measured by the same method as every other brand, with no adjustment. None of them was named in this category.

9The questions

Each question links to its page, with the four answers in full, the brands read from them and the pages they cited.

10Data

• Every answer: all 240 in full, with the searches each AI ran, the sites it found and cited, and the brands readCSVJSON
• All 372 brands, every measureCSVJSON
• The reportPDF

Licence: CC BY 4.0. Use and share with credit to Citewake.

11How to cite

Citewake. (2026). AI writing assistants: what AI recommends, November 2026 (Edition 1) [Data set]. https://doi.org/10.5281/zenodo.23763253

@techreport{citewake202611aiwritingassistants,
  author      = {{Citewake}},
  title       = {{AI writing assistants: what AI recommends, November 2026}},
  institution = {Citewake},
  type        = {Edition 1, data set},
  year        = {2026},
  month       = oct,
  version     = {1},
  doi         = {10.5281/zenodo.23763253},
  url         = {https://citewake.com/editions/2026-11/ai-writing-assistants},
  note        = {Licensed CC BY 4.0}
}