Short answer: A popular generative engine optimization tip says you should load your writing with first-person experience (“I tested,” “in my experience”) because AI engines favor it. We tested that claim. Across 56 editorial pages, we measured the density of first-person language and compared the pages Google’s AI Overview cited against the ones it ignored. The cited pages used slightly less first-person language, not more (1.47 versus 1.73 first-person words per 100 words). The “write more in first person and AI will cite you” advice did not hold up. We call it the first-person fallacy.
One of the most repeated pieces of generative engine optimization advice goes like this: AI engines crave human experience to compensate for not having any, so stuff your posts with first-person, experiential language and you will get cited more. It sounds plausible. It is also testable. So we tested it, and the data says the advice is, at best, not doing what people claim.
What the first-person citation claim says
The theory is that phrases like “I tested,” “we found,” and “in my experience” signal the first-hand expertise that Google’s E-E-A-T guidelines reward, and that AI answer engines therefore preferentially quote passages dense with first-person pronouns. Writers have been told to rewrite neutral prose into personal testimony specifically to win AI citations. We wanted to know whether the pages that actually get cited really are more first-person than the ones that do not.
How we tested it
We took 56 editorial and blog pages that appeared in Google’s organic results for our blogging and SEO queries, and split them into two groups: the 27 that Google’s AI Overview cited for that query, and the 29 it did not. For every page we measured first-person pronoun density: the count of words like I, me, my, we, us, and our, plus their contractions, per 100 words of text. We deliberately excluded forums and social platforms like Reddit and Quora, because those are first-person by nature and would swamp the comparison. This is a clean test of editorial prose against editorial prose.
What we found: no first-person advantage
If the theory were right, the cited group would show clearly higher first-person density. It did not. If anything, the relationship ran the other way.
- Cited pages: 1.47 first-person words per 100 words on average (median 1.09).
- Non-cited pages: 1.73 per 100 words on average (median 1.43).
- Pages with heavy first-person density (2 or more per 100 words) made up 33 percent of cited pages versus 38 percent of non-cited pages.
On every measure, the pages Google’s AI Overview chose to cite were slightly less first-person than the ones it passed over. The gap is small and we would not claim first-person language actively hurts you. But the confident claim that piling on “I tested” language earns AI citations is simply not visible in the data. We call this the first-person fallacy.
Why the advice misfires
The likely confusion is a real signal mistaken for the wrong cause. Forums like Reddit do get cited heavily by AI Overviews, and forums are written entirely in the first person. So it is easy to conclude that first-person voice is what earns the citation. But the citation is going to Reddit because it is Reddit, a massive, trusted, experience-rich platform, not because a comment said “I.” When you compare editorial pages to editorial pages, holding platform authority roughly constant, the voice difference disappears.
In other words, what looks like a writing-style signal is really a source-authority signal. We measured that authority effect directly in our companion study on who AI Overviews cite, where brand and forum sources dominate. Rewriting your prose into first person does not turn your blog into Reddit.
What actually correlates with citation
If voice is not the lever, what is? Our broader research points to three things that do track with getting cited: being a recognized or primary source, stating facts in clean and extractable sentences, and putting the answer where a model can reach it rather than burying it. That last point is measured in our answer depth study, and you can pressure-test any page against these signals with our AI Citation Grader.
None of this means you should strip the human voice out of your writing. First-person experience builds trust with readers, and trust drives the behavior that does matter. Just do not write it because you think the pronouns themselves buy you AI citations. On our evidence, they do not.
How we measured this (methodology)
In June 2026 we identified editorial and blog pages appearing in Google’s organic results for 30 blogging and SEO queries, and labeled each as cited or not cited based on whether Google’s AI Overview for that query linked its domain. We fetched each page and counted first-person pronouns (I, me, my, mine, myself, we, us, our, ours, ourselves, and contractions) per 100 words of page text. Forums, Q&A sites, and social platforms were excluded so the comparison reflects editorial writing rather than inherently personal user content. Four pages were dropped for returning no readable body, leaving 56 (27 cited, 29 not). Density was computed on full page text, which adds a small uniform amount of boilerplate to every page and does not favor either group. The dataset is published below.
Frequently asked questions
Does writing in first person help you get cited by AI?
Not in our data. Across 56 editorial pages, the ones Google’s AI Overview cited used slightly less first-person language (1.47 per 100 words) than the ones it ignored (1.73). We found no citation advantage from first-person density.
But forums full of “I” get cited constantly. Why?
Because of what they are, not how they are written. Reddit and similar platforms are cited for their scale and trusted user experience. When you compare editorial pages to editorial pages, the first-person effect disappears, which suggests the real signal is source authority, not voice.
Should I remove first-person language from my posts?
No. First-person experience builds reader trust and supports E-E-A-T for human readers. Just do not add it specifically expecting it to earn AI citations, because our data does not support that.
What does correlate with AI citation?
Being a recognized or primary source, stating facts in clean extractable sentences, and positioning the answer near the top rather than burying it. Original data of your own is one of the strongest signals.
Original study by Blogging Titan. Published June 2026. Related first-party research: who AI Overviews cite and the rank versus citation disconnect.
This is original first-party research by Blogging Titan. The dataset below is free to cite or republish with attribution under a CC BY 4.0 license.
| Editorial pages analyzed | 56 (27 cited, 29 not) |
| First-person density, cited pages | 1.47 / 100 words |
| First-person density, non-cited pages | 1.73 / 100 words |
| Citation advantage from first person | None found |
| Heavy first-person (2+/100w): cited vs not | 33% vs 38% |
Blogging Titan. (2026). The First-Person Fallacy: First-Person Language and AI Citations. Retrieved from https://bloggingtitan.com/blog-seo/first-person-language-ai-citations-study/
@misc{bloggingtitan_first_person_fallacy_2026,
title = {The First-Person Fallacy: First-Person Language and AI Citations},
author = {{Blogging Titan}},
year = {2026},
url = {https://bloggingtitan.com/blog-seo/first-person-language-ai-citations-study/},
note = {Original first-party dataset, CC BY 4.0}
}
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.
