Blog SEO

AI Citations per Answer: Conductor’s US Benchmarks vs Our Blogging Sample (31 October 2026)

The comparison table

Each row puts Conductor’s published figure next to the closest thing we can measure on blogging searches. The right-hand column says how far the two can fairly be compared, because some of these are like for like and some are not.

Conductor’s figures come from its 2026 Publishing AI search benchmarks, read on 31 October 2026 (the page shows a last update of 26 August 2026). The answer-engine figures cover 1 January to 31 May 2026 across seven engines and more than 690,000 prompts. The AI Overview figures cover 20 May to 20 June 2026 and 16.1 million US Google searches. Citations count as publishing because of the industry of the cited domain, not the topic of the question.

What we measured, and on which day

The 30 queries are the fixed set from our Reddit Tax census: twelve informational searches such as “how to start a blog”, ten “best” searches such as “best keyword research tool” and eight product reviews such as “semrush review”. Using the same set lets later readings be compared with earlier ones.

On 31 October 2026 we loaded each query in Chrome with Google set to the United States and English. For the AI Overview we opened “Show more” and recorded every source link inside the overview. For the organic top 10 we took the first ten results with a title link in the main results column, reading page two where page one ran short. Video carousels, “People also ask” and other features do not count as organic positions. For AI Mode we opened the same query in AI Mode and read the citation chips once the answer had finished.

The ChatGPT and Perplexity rows come from our AI Citation League Table, which asked each engine 100 blogging questions on 2 September 2026. ChatGPT was signed in on the free plan and Perplexity was signed out.

AI Mode: about 7 sources, if you count the names

AI Mode attaches small chips to its sentences. Each chip names one source and often adds a count, for example “YouTube · Nakisha Wynn (+2)”, meaning two more sources sit behind it. Those extra sources are not in the page until each chip is opened, and opening one follows the link. That leaves two honest counts:

• Distinct sources named on the chips: 6.4 per answer on average, median 6, across 23 answers. This is a floor, since a hidden source can be a site the chips never name.
• Chip attributions including the “+N” sources: 22.0 per answer, median 22. This is a ceiling, since the same source is often attached to several sentences.

Conductor’s 7 sits just above our floor. For a blogger the floor is the useful number: on a typical blogging answer, about six names get shown, and the reader sees them without opening anything.

YouTube led 78 of the 204 chips (38.2 percent) and appeared on 18 of the 23 answers.

ChatGPT and Perplexity: the averages hide the shape

Conductor reports 3.9 citations per answer for both. Our September reading looks nothing like that, and the reason is how each engine behaves on a blogging question.

ChatGPT cited no source on 48 of 100 blogging questions (two of those answers carried a plain link in the text). Averaged across every answer it gave 2.44 citations. Averaged across the 52 answers that cited something, it gave 4.69, above Conductor’s figure. Conductor does not say whether its average includes answers without citations, and that choice alone moves the number a long way.

Perplexity, signed out, returned exactly ten sources on 99 of 100 answers, for 9.99 per answer. A fixed ten-source list looks like a retrieval setting. A count taken from a signed-in session, a different plan, or only the sources referenced inline may well come out lower. Treat our figure as a measure of what a signed-out reader is shown.

AI Overviews: the outside-the-top-10 share is mostly video

Conductor found that 57.8 percent of publishing AI Overview citations came from pages outside the organic top 10. On blogging searches we found almost the same share: 159 of 266 links, or 59.8 percent. At the level of the whole domain the share falls to 48.5 percent (129 of 266), because some cited pages sit on a site that already ranks with a different page.

The breakdown changes the story. YouTube supplied 108 of the 266 links, and 105 of those 108 were outside the organic top 10. YouTube appeared in every one of the 29 overviews we could read, yet took only 8 of the 290 organic top-10 positions. With YouTube left out, 54 of the remaining 158 links came from outside the top 10, which is 34.2 percent.

So for a text page on a blogging topic, the old rule mostly still holds. About two in three of the non-video links in these overviews also ranked in the top 10 for that query. The “you can be cited without ranking” story is real, and on these searches it is mostly a video story.

By intent, the outside share was 54.1 percent on informational searches (60 of 111 links), 70.8 percent on “best” searches (51 of 72) and 57.8 percent on reviews (48 of 83).

Our earlier rank-citation disconnect study put the figure at 68 percent in July. That reading used a different query mix and captured the organic list differently, so the two are not a trend line.

Position 1 was cited more often on our searches

Conductor says a page at organic position 1 made the AI Overview only 40.7 percent of the time. On our blogging searches the number-one page was cited on 17 of 29 queries, or 58.6 percent. Review searches were the strongest (6 of 8, four of them Trustpilot pages) and “best” searches the weakest (3 of 10).

What this means for a small blog

• Write for about six visible slots in AI Mode. The answer may draw on more, but roughly six names reach the reader on a blogging question.
• A video can be cited from outside the top 10. On these searches YouTube took 41 percent of AI Overview links while holding under 3 percent of organic positions. If a topic is visual, a short video on the same subject is the route that most often bypasses ranking.
• For a text page, ranking and citation still go together. About two thirds of non-video AI Overview links came from pages already in the top 10.
• Read engine averages with the definition attached. ChatGPT’s 3.9 and our 2.44 measure different things; across only the answers that cited something, ours was 4.69.

Limits

Small sample: 30 queries read once on one day, from a UK Chrome with US Google settings, against Conductor’s 690,000-plus prompts and 16.1 million US searches across 672 publishing domains. Seven of the 30 AI Mode answers are missing: Google’s automated-traffic check stopped the run before six of them, all review searches, and one answer showed no citation chips when read. The AI Mode figures therefore cover only 2 of the 8 review searches. One AI Overview (“how to grow a blog fast”) rendered in a layout whose sources were not in the page as links. It is left out of the overview counts, leaving 29. Overview counts include only links in the expanded overview, not sources hidden behind “+N” groups. AI Mode counts come from chip labels, as described above. ChatGPT and Perplexity figures are from 2 September 2026 and a different 100-question set. Treat every number here as a dated spot check to re-run, not a benchmark.

@misc{bloggingtitan_citations_per_answer_2026,
  title  = {AI Citations per Answer: Conductor's US Benchmarks vs Our Blogging Sample (31 October 2026)},
  author = {{Blogging Titan}},
  year   = {2026},
  url    = {https://bloggingtitan.com/blog-seo/ai-citations-per-answer-benchmark/},
  note   = {Original first-party dataset, CC BY 4.0}
}

Sources

• Conductor, 2026 Publishing AI search benchmarks (Jia-Rong Li; last updated 26 August 2026; read 31 October 2026)
• Blogging Titan, The AI Citation League Table (data collected 2 September 2026)

Related guides

• The AI Overview citation appetite: how many sources Google’s AI cites
• Reddit ranks on all 30 blogging searches; Google’s AI quotes it on 18
• The rank-citation disconnect
• Which UK businesses does AI recommend? What 160 Google AI Mode answers show

Written by

Blogging Titan Team

Blogging Titan is an independent team of bloggers documenting what actually grows a blog in the AI search era. We have been building, ranking, and monetizing WordPress sites since 2017, and every guide on this site is based on strategies and tools we have tested ourselves. Want a second pair of eyes on your blog? Request a free blog audit or start with the 2026 playbook.

Summarize with: ChatGPT Perplexity Claude Grok