Key takeaways
- From Sep 28 to Oct 4, 2026, social platforms made up 0.46% of ChatGPT's citations and 12.2% of Google AI Overviews' citations, a gap of roughly 26x, according to Promptwatch's social citations report.
- Most of ChatGPT's drop is Reddit. Its share fell from about 4% to about 0.5% on August 14, 2026.
- YouTube leads in AI Overviews (4.28%), with Facebook, Reddit, Instagram and LinkedIn behind it. Perplexity sits in the middle at 2.7%.
- Profound's March 2026 study found a different split (ChatGPT 9.1%, AI Overviews 15.3%). The numbers differ because of timing and methodology, not because one study is wrong.
- Plan social presence per engine. One blanket strategy no longer works.
The headline numbers
The gap between ChatGPT and Google's AI Overviews is the biggest single difference in citation behavior I have seen this year. Same kind of question, same kind of user, wildly different source mix.
Promptwatch Data tracked seven social platforms (YouTube, Reddit, LinkedIn, Facebook, Instagram, TikTok and X) across three engines for the week of Sep 28 to Oct 4, 2026. Each figure below is that platform's share of all of that engine's citations.
| Engine | Social share of citations | Top social platform |
|---|---|---|
| ChatGPT | 0.46% | LinkedIn (0.3%) |
| Perplexity | 2.7% | LinkedIn and Reddit (0.86% each) |
| Google AI Overviews | 12.2% | YouTube (4.28%) |
Averaged across the three engines, YouTube comes out at 1.70%, Reddit at 1.03%, LinkedIn at 0.90%, Facebook at 0.81%, Instagram at 0.56%, TikTok at 0.07% and X at 0.04%. X barely registers anywhere.
One caution before you quote any of this: these are shares of citations in one week. They move. Treat them as a snapshot of a shifting system, not a permanent ranking.
Why ChatGPT dropped to 0.46%
The short answer is Reddit. The longer answer is that the change happened in a few days and nobody outside OpenAI knows exactly why.
In the earlier edition of Promptwatch's social citations report, ChatGPT drew 5.8% of its citations from social platforms. Now it is 0.46%. Reddit alone went from 5.19% to 0.09%.
Promptwatch's Reddit citation data shows the timeline:
- Reddit held a steady 3.83% share of ChatGPT citations from Jul 18 to Aug 7, 2026.
- On Aug 8, the day ChatGPT changed its query fan-out behavior, Reddit slipped to the mid-2s.
- On Aug 14 it fell below 1%. The Aug 14 to 17 average was 0.52%, an 86.4% relative drop.
The Aug 8 date lines up with another shift: ChatGPT started using the site: operator at scale, jumping from about 0.4% to about 17% of fan-out queries overnight. Promptwatch is careful here, and so should we be. The chart shows when things changed, not why. A change in source selection is the obvious candidate, but a data-collection issue can't be fully ruled out, so the size of the drop is provisional.
What I find notable is how little replaced it. Social did not simply shift to another platform inside ChatGPT. LinkedIn is now the top social source at 0.3%, which is tiny. ChatGPT mostly moved toward other content types. Promptwatch's content type data puts product pages first at 22.4% of citations, followed by landing pages and listicles.
That matches the pattern described in the Everything-PR citation index, which argues that ChatGPT has leaned toward extractable, authority-tagged content and away from open-forum content since late 2025. That is one analyst's reading, but it fits the numbers.
Why AI Overviews cite social at 12.2%
Google's AI Overviews behave almost the opposite way. Social platforms are 12.2% of their citations, about 26 times ChatGPT's share. The platform mix:
| Platform | Share of AI Overview citations |
|---|---|
| YouTube | 4.28% |
| 2.38% | |
| 2.15% | |
| 1.61% | |
| 1.55% | |
| TikTok | 0.22% |
Two details stand out. Facebook more than doubled and passed Reddit. And TikTok shows up only in AI Overviews. Google owns YouTube and indexes these platforms heavily, so it is not shocking that its own answer surface leans on them. It is still a big contrast with ChatGPT.
Reddit in AI Overviews did decline, but gently. Promptwatch measured it falling from about 2.5% in early July to about 2.1% in August, an 11.3% relative decline. In AI Mode the drop was 30.5% (2.22% to 1.54%), with the clearest step at the end of July. Gradual erosion, not a cliff.
Perplexity: the middle child
Perplexity cites social at 2.7%. LinkedIn and Reddit tie at 0.86%, and YouTube is at 0.79%, down from 2.67% in the earlier report. That YouTube drop is the sharpest on Perplexity and worth watching if video is part of your plan. X peaks at around 0.1%.
What the other studies say
Promptwatch's numbers are not the only ones. They also differ from them, and the differences are instructive.
Profound analyzed 3.25 billion citations across 7 models and 14 countries in March 2026. Its social citation rates by model:
| Model | Social citation rate (Profound, March 2026) |
|---|---|
| Google AI Overviews | 15.3% |
| Google AI Mode | 14.5% |
| Perplexity | 11.3% |
| ChatGPT | 9.1% |
| Copilot | 4.3% |
| Claude | 3.99% |
| Gemini | 3.6% |
Why the numbers don't match
Don't average the two datasets or call one wrong. Three things explain the gap:
- Timing. Profound's data is from March 2026, before ChatGPT's Reddit collapse in August. A 9.1% ChatGPT figure in March and 0.46% in early October is consistent with that shift.
- Scope. Profound's study is international and filtered each prompt by the country's native language. Promptwatch's report covers a single week of tracked prompts.
- Classification. Profound's social bucket may be broader than the seven platforms Promptwatch counts.
The direction is the same in both: AI Overviews cite social the most, and Google's two surfaces sit well above ChatGPT. Even within one company, models differ. Profound found 15.3% in AI Overviews versus 3.6% in Gemini.
Language changes the picture
Profound also found that query language matters more than country. In AI Overviews, Mexico showed a 22.5% social citation rate and Spain 20.0%, while the same engine nearly erased social citations in Swedish and Arabic. If you operate in several markets, one global average hides real differences. You need per-language numbers.
Different engines, different sources
Social share is one slice of a bigger divergence. SISTRIX analyzed 82,619 prompts over 17 weeks and found that AI Overviews and AI Mode cite different domains 83% of the time for the same prompt. Overlap between AI Mode and ChatGPT is even lower.
The practical read: winning a citation in one engine tells you little about the others. Check each one separately. Our own data backs this up. See the broader numbers in Promptwatch Data.
What to do with this
If your buyers use Google
Invest in YouTube first. It is the largest social source in AI Overviews at 4.28%. After that, keep public Facebook pages and Instagram profiles complete and active, since both are now meaningful in AI Overviews.
If your buyers use ChatGPT
Don't count on Reddit threads to get you in. Promptwatch's recommendation is to put that effort into your own site and the listicles, guides and comparisons ChatGPT does cite. Page position matters too: CXL's analysis of 100 AI Overview citations found 55% of cited passages in the first 30% of the page, and Kevin Indig found 44.2% across 18,012 ChatGPT citations. Put the answer early.
If Perplexity matters to you
Publish on LinkedIn as well as Reddit, since they tie at 0.86%.
What to skip
Leave X out of AI search plans. At 0.04% across engines it is not worth the effort for citation purposes.
How to track this yourself
You can't manage a per-engine strategy from one blended number. You need citation data by engine, by platform and over time.
Promptwatch tracks real-UI responses from ChatGPT, Google AI Overviews, AI Mode, Perplexity and others, and has dedicated Reddit, YouTube and LinkedIn citation reports, with a "My channels" filter for your own accounts.
Other options worth a look, depending on your needs:
| Tool | Good for |
|---|---|
| Promptwatch | Citation tracking by engine and platform, plus content and crawler data to act on it |
| Profound | Enterprise-scale monitoring and research |
| Peec AI | Multi-language tracking |
| Otterly.AI | Lightweight monitoring on a budget |
If you want to compare more options, the AI rank tracking directory lists many of them.
A sensible testing routine
- Pick 30 to 50 prompts your buyers actually ask. Include some brand-specific and some comparison prompts.
- Track them in ChatGPT, AI Overviews, AI Mode and Perplexity separately.
- Log which social domains appear as citations for each engine, weekly.
- Compare before and after any big engine change. The Aug 8 and Aug 14 shifts would have been visible within days.
- Shift effort toward the platforms that actually earn citations in the engine your audience uses.
Caveats
Social share of citations is not the same as influence on answers. A model can read a Reddit thread to form an answer and never cite it. Promptwatch also notes the August drop could include data-collection effects. And a week of data is a week of data. Re-check monthly.
The core finding stands, though: ChatGPT now almost ignores social platforms as cited sources, Google's AI Overviews lean on them heavily, and Perplexity sits between. Build for each engine on its own terms.