OtterlyAI’s September 2026 data, published September 22 and drawn from 1,600 US prompts run daily across seven AI engines, shows LinkedIn’s share of AI citations rising 39% within the month and YouTube’s falling 42%. Reddit was not halved in September: across all engines its share rose about 10% through the month. The Reddit collapse is real, but it is narrower and older than the headline suggests: inside ChatGPT, Reddit fell from 3.765% of citations in early summer to 0.227% in September, a 94% drop that began on August 8. Almost all of YouTube’s fall came from one engine, Google AI Mode, which cut YouTube’s share of its citations by about 72% in 48 hours. The lesson is not “move your budget to LinkedIn.” Each engine re-weights its sources on its own schedule, so brands need presence across several source types and tracking that shows which engine moved.
If you have seen September summarized as “LinkedIn up, YouTube down, Reddit halved,” two of the three numbers hold up. The third needs a correction, and the corrections tell you more than the headline does.
What the September data actually shows
The original source is OtterlyAI’s “September Update,” written by Rick Tousseyn. It uses Otterly’s AI Search Index: 1,600 prompts across 16 US industry reports, run daily against ChatGPT, Google AI Mode, Google AI Overviews, Claude, Perplexity, Gemini and Microsoft Copilot. Every figure is the median of daily shares, meaning the percentage of tracked citations on a given day that pointed to a domain. Otterly does not publish a total citation count for the period.
LinkedIn went from 0.407% of all citations on September 1 to 0.564% on September 20. It gained on five of seven engines, led by ChatGPT (+66%), Google AI Mode (+62%) and Perplexity (+51%). Google AI Overviews moved the other way (-13%), and Gemini did not cite LinkedIn once in the September window.
YouTube went from 2.221% to 1.292% over the same dates. Otterly describes the fall as “a single step around 13 to 15 September rather than a drift.”
Two caveats belong next to those headline numbers. First, they are within-month trajectories. Month over month, August to September, all three platforms lost share: LinkedIn -4%, YouTube -16%, Reddit -3%. Second, the bases are small. LinkedIn at 0.564% is roughly one citation in 177. Social platforms as a group took 4.311% of citations in September, down from 5.146% in August and 5.983% in July, while brand-owned pages took about 58% in Otterly’s category table. (That table puts social at 6.632%, a different figure from the 4.311% above; the article does not reconcile the two, which is a reason to read any single number from it with care.)
Reddit was not halved in September. In ChatGPT, it was cut by 94%
Across all seven engines, Reddit’s share rose from a 1.597% median in the first ten days of September to 1.751% in the last ten. Underneath, the engines pulled in opposite directions: Google AI Overviews (+10%) and Google AI Mode (+22%) cited Reddit more, Perplexity (-11%) and Gemini (-9%) cited it less. Claude did not cite reddit.com on any of the 77 days Otterly measured.
The real Reddit story is ChatGPT. Reddit held 3.765% of ChatGPT’s citations from July 1 to August 7, 0.553% from August 8 to 24, and 0.227% in September. That is the 94% drop, and it is the one change of the quarter that has not reversed.
An earlier Otterly study, published August 27, compared August 6 to 13 with August 14 to 17 across 16 brand reports: ChatGPT’s daily Reddit citations fell from an average of 497 to 132 (-73.4%, which Otterly calls a conservative floor while flagging the short four-day post window as a limit), while ChatGPT’s own citation volume rose 3.5% and Reddit citations on four other engines held flat (-0.6%). Gizmodo, citing PromptWatch data on August 18, reported that Reddit had appeared in as many as 4.5% of ChatGPT outputs until August 8 and in about 0.5% afterward. Two vendors, different denominators, same direction.
So if “Reddit halved” came from somewhere, the closest real figures describe ChatGPT in August, not every engine in September. Reddit still matters for Google’s AI surfaces and Perplexity, which is where the tactics in our guide to getting cited on Reddit still pay off.
Why citation sources shift
Citation share moves for reasons that mostly sit on the engine’s side, not yours.
Retrieval changes inside one engine. The ChatGPT Reddit drop landed in discrete steps on August 8 and August 14, and other engines kept citing the same Reddit threads. Otterly’s conclusion: “this was a ChatGPT decision rather than anything Reddit did.” PromptWatch, via Gizmodo, said OpenAI began using a “site:” operator when searching, going to specific sites instead of starting from the open web. Otterly labels that the leading hypothesis, not a conclusion, and OpenAI had not responded to Gizmodo by publication.
Surface changes inside Google. Google AI Mode cited YouTube for 11.0% of its citations on September 13, 6.8% on September 14 and 3.1% on September 15, then held between 3.4% and 3.9%. Otterly does not identify a cause. It had built that position over the summer, from 4.2% in July to a 14.1% peak on August 29, and gave it all back in the two days from September 13 to 15.
Licensing deals that do not guarantee citations. OpenAI and Reddit announced a partnership on May 16, 2024, giving OpenAI access to Reddit’s Data API, described as “real-time, structured, and unique content from Reddit.” ChatGPT still cut Reddit citations in 2026. A content deal is not a citation guarantee.
Volume swings that look like trends. Otterly notes that ChatGPT’s citation volume recovered 65% in the first half of September, which “explains most of its share movements.” When the denominator moves, every share moves with it.
The broader mechanics of how engines pick sources are covered in where AI actually gets its answers, and the run-to-run side of the same problem in why AI recommendations change overnight.
Why one month’s swing is not a strategy
The September data is useful. It is also a poor basis for reallocating a content budget, for four reasons.
Moves reverse. ChatGPT’s LinkedIn share went from 0.116% to 0.551% after August 8, then fell back to 0.231% in September. Otterly’s own summary: “ChatGPT is where the quarter’s biggest swings happened, and September walked most of them back.” Perplexity’s LinkedIn share stepped down to 0.57% between July 29 and August 2 and jumped back to 1.10% on September 8.
Percentages on tiny bases mislead. Perplexity’s YouTube citations rose 250%, from 0.056% to 0.196%. Otterly itself calls that “a real direction” but “not a meaningful volume.”
Engines disagree. Gemini barely cites LinkedIn: zero on 45 of the 77 days Otterly measured, and never above 0.03%. Claude did not cite Reddit on any of those days. ChatGPT barely cites Reddit now. A “social” or “Reddit” strategy that assumes one behavior across engines is wrong about some of them by design.
The sample has limits. This is US data, 16 industries, a September window that ends on September 20, and five days of data after the AI Mode step. Your category and market may behave differently. Otterly’s August study put it plainly: “Do not rebuild your strategy around a four-day window.”
Longer trends deserve more weight. Otterly’s June study of 1,310,455 LinkedIn citations found LinkedIn’s share of social citations rising from 7.8% in January 2026 to 11.7% in May. Five months of direction is a better planning input than 20 days.
How to spread your presence and track sources over time
The goal is a source mix that no single engine update can wipe out, plus a measurement loop that tells you which engine moved before your pipeline does.
-
Build the owned base first. Brand-owned pages took about 58% of September citations in Otterly’s category table. After ChatGPT’s August change, official brand and competitor pages rose among ChatGPT’s top-cited domains in 10 of the 15 Otterly brand reports where they appear. Docs, help center, pricing and comparison pages are the layer you control.
-
Publish long-form LinkedIn articles under named people. In Otterly’s LinkedIn study, Pulse articles took 72.2% of LinkedIn content citations, named individuals took 91.7%, and likes and comments correlated near zero with citations. Perplexity accounted for 43.3% of LinkedIn citations in that dataset. Expect almost nothing from Gemini, which cited LinkedIn 38 times in it.
-
Make YouTube long-form and chaptered. Long-form videos drew 94% of YouTube citations in Otterly’s March study, and every timestamped citation it observed came from Google AI Overviews (73%) or AI Mode (27%). After the September step, Otterly says Google AI Overviews is the only engine still citing YouTube at scale.
-
Treat Reddit as a Google and Perplexity channel. Participate for real, in threads your buyers already read. Do not count on it for ChatGPT or Claude visibility right now.
-
Earn trade press and review coverage. News and media took 15.772% of September citations in Otterly’s table. The studies above do not break out review sites, so check which ones your engines cite in your own category.
-
Track citations per engine, on a fixed prompt set, every week. Split by engine and by cited domain, and keep each engine’s total citation volume next to the shares. Otterly’s August study showed why: across all engines Reddit fell only 14.4%, which hid a 73% drop inside ChatGPT. The setup is covered in how to track citations and sources in AI search.
-
Set a concentration ceiling and a waiting period. If one domain on one engine carries most of your citations, you hold the same risk Reddit held in ChatGPT. Act on a shift only when it persists for several weeks in your own prompts.
What to watch next
The useful question after September is not “which platform is winning” but “which engine changed, for which of my prompts, and did it stick.” That question needs the same prompts run the same way over time, with sources broken out by engine. Rankry runs a stable set of buyer-intent prompts every day across ChatGPT, Claude, Gemini, Perplexity, Grok and Google AI Mode and AI Overviews, and shows how each engine mentions, ranks, cites and recommends your brand. When an engine re-weights its sources, you see it as your own line moving, engine by engine, and competitor comparison shows whether the same shift hit everyone in your category or only you.
FAQ
Did Reddit’s AI citations really halve in September 2026? Not across AI search as a whole. In OtterlyAI’s data, Reddit’s share across seven engines rose about 10% through September and fell only 3% from August. The large drop is specific to ChatGPT, where Reddit went from 3.765% of citations in early summer to 0.227% in September, a 94% decline that started on August 8.
Why did YouTube citations fall 42% in September 2026? Almost entirely because of Google AI Mode. Its YouTube share fell from 11.0% on September 13 to 3.1% on September 15 and stayed between 3.4% and 3.9% afterward. OtterlyAI did not identify a cause. Google AI Overviews declined more gently, while several smaller engines cited YouTube slightly more from very small bases.
Should I move my content budget to LinkedIn? Not on one month of data. LinkedIn’s 39% rise was within September; month over month it fell 4%, and it is still about half a percent of all AI citations. The longer trend is more persuasive: its share of social citations grew from 7.8% to 11.7% between January and May 2026. Add LinkedIn articles to a mix, not as a replacement.
Which AI engines ignore LinkedIn, Reddit or YouTube? In OtterlyAI’s September data, Gemini did not cite LinkedIn at all, Claude did not cite reddit.com on any of 77 measured days, and Microsoft Copilot cited Reddit rarely and never above 0.05%. ChatGPT now cites Reddit at a fraction of its July level. After mid-September, Google AI Overviews was the only engine still citing YouTube at scale.
How do I track which sources AI engines cite for my brand? Run a fixed set of buyer-intent prompts on a regular schedule, record every cited domain, and split the results by engine. Keep each engine’s total citation volume next to the shares, so a volume swing is not mistaken for a trend. Compare week over week and act only on shifts that persist for several weeks in your own prompts.
Sources
- The September Update: The Month AI Search Stopped Agreeing About Social, OtterlyAI, September 22, 2026
- ChatGPT cut Reddit citations by at least 73% in August 2026. Here is what replaced it., OtterlyAI, August 27, 2026
- OpenAI Is Backing Away From Reddit as Reddit Tries to Become OpenAI?, Gizmodo, August 18, 2026
- OpenAI and Reddit Partnership, OpenAI, May 16, 2024
- 1 in 8 Social Media AI Citations Point to LinkedIn: AI Search Study by OtterlyAI, OtterlyAI, June 3, 2026
- The YouTube Citation Study 2026, OtterlyAI, March 2, 2026
See which sources each AI engine cites for your buyer prompts, every day, and catch the next re-weighting as it happens. Start a free 7-day Rankry trial, no card required.