Summary. This experiment changed nothing on the page except the date, and the company that ran it sells the tool that measured the result. AI-visibility vendor OtterlyAI changed only the visible "Last updated" date on 10 of its own blog articles on 2026-08-04 and left every word of the content alone. Its tracker counted 4,800 AI citations to those articles in the 30 days before the change and 6,312 in the 30 days after, a rise of 31.5%. Two articles left untouched as controls fell from 358 to 320 citations over the same windows. OtterlyAI published the experiment on 2026-09-24 and calls its own result "directional rather than conclusive", because citation trends were already moving before the date changed.
Credibility: MEDIUM. We read OtterlyAI's write-up directly. No independent outlet has replicated or corroborated it. OtterlyAI sells AI-visibility tracking, the test used 10 treated pages and 2 control pages, and the author's own caveat applies.
Facts
- Method: visible dates changed on 10 articles on 2026-08-04; 2 articles left unchanged as controls. Before window 2026-07-05 to 2026-08-03; after window 2026-08-05 to 2026-09-03.
- Engines tracked: ChatGPT, Claude, Perplexity, Google AI Overviews, Google AI Mode, Gemini and Copilot.
- Result: citations to the 10 updated articles rose from 4,800 to 6,312 (+31.5%). The 2 controls fell from 358 to 320 (-10.6%). 7 of the 10 updated articles gained citations, 1 was flat and 2 declined.
- Per-engine and secondary figures reported by OtterlyAI: ChatGPT +692%, Perplexity +217%, Claude +87%. Google organic clicks to the updated articles rose from 80 to 132 (+65%).
- Stated limitation: the citation trend was already rising before the date change.
Sources
OtterlyAI Blog: We Updated Only the Publication Date on 10 Articles. AI Citations Rose 31.5%, OtterlyAI (Thaylise Nakamoto), 2026-09-24, https://otterly.ai/blog/publication-date-ai-citations-experiment/
Stay ahead of AI search
The changes that move your rankings and AI citations — and the exact move to make — before they cost you. Free, three times a week.
For a marketing lead deciding what a content-refresh budget should buy, this is a hypothesis with a good headline, and OtterlyAI deserves credit for printing the caveat itself. Ten treated pages, two controls and a trend that was already rising leave three explanations standing: the new date, the existing trend, or ordinary movement between two 30-day tracker windows. The experiment cannot rank them. The per-engine jumps (ChatGPT +692%) are the figures most likely to get forwarded around your company, and the ones that deserve the most caution, since they come from the same 10 pages. Suppose the date does move citations. The practical question is still whether the date on the page tells the truth. A "last updated" stamp on an article nobody updated is a claim a reader can check, and when it fails the check, the brand is the one that pays. There are two ways to get this wrong. One is to ignore a cheap, testable lead. The other is to re-date a whole site on the strength of 10 pages. A tracker is the right tool for spotting movement like this; a controlled test on your own pages is what turns the movement into a decision.
Actions this week
1. Ask your web team to list which pages show a "last updated" date, and whether the CMS ties that date to a real content edit or to a field anyone can change by hand.
2. Within 30 days, run your own controlled test: pick 10 comparable pages, make a genuine refresh of the facts on all 10, change the visible date on only 5, and compare citation counts in your AI-visibility tracker against Search Console clicks over 30 days.
3. Until that result is in, keep date-only changes (a new date, no new content) out of the standard refresh deliverable, and write down why in the client file.
OtterlyAI Blog: We Updated Only the Publication Date on 10 Articles. AI Citations Rose 31.5%, OtterlyAI (Thaylise Nakamoto), 2026-09-24. https://otterly.ai/blog/publication-date-ai-citations-experiment/