What Actually Moves ChatGPT Citations: Lessons From Promptwatch's Content-Type and Domain-Rank Data (2026)

Key takeaways

• Product pages (22.4%) and landing pages (16.8%) together earn 39% of ChatGPT citations, per Promptwatch Data for Sep 28 to Oct 4, 2026.
• Domains ranked DR 46-75 earned about 46% of citations in August 2026. DR 91-100 earned 4.4%. You do not need a giant domain.
• Content type decides how long a citation lasts: landing pages and docs hold on, news loses half its pages in 36 days.
• Older "Reddit is 16.8% of ChatGPT citations" figures are out of date. Promptwatch saw Reddit's share fall from about 4% to 0.5% on August 14, 2026.
• Track citations per content type and per engine on a locked prompt list, then fix the one gap your data shows.

Why most citation advice is a bit off

Read five guides on getting cited by ChatGPT and you get the same eight-factor lists. Domain authority, schema, freshness, expert bylines, and so on. None of it is wrong exactly. But it is rarely ranked by what actually shows up in citation data, and some of it was true of a ChatGPT that no longer exists.

Take the claim that Reddit and Wikipedia dominate, with Reddit at roughly 16.8% of citations. That number comes from a September 2026 third-party ranking. Promptwatch's own tracking shows reddit.com's share of ChatGPT Search citations holding near 3.8% through early August, then falling to about 0.5% on August 14, 2026. Both can be partly true depending on the sample window and method, but if you build strategy on the bigger number you will be planning for a source mix that has already shifted. See Promptwatch's report on Reddit citations dropping in ChatGPT for the daily chart.

This guide leans on two first-party datasets from Promptwatch: citation share by content type, and citation share by domain rank. Plus the content half-life study. Together they answer three practical questions. What kind of page gets cited? How big does your domain need to be? How long does a citation last?

What type of page gets cited

Promptwatch Data classified ChatGPT citations from Sep 28 to Oct 4, 2026 by content type. The full report is here. The headline split:

Two things jump out. First, commercial pages beat editorial ones. Product and landing pages are 39% of everything cited. If your content plan is 90% blog posts, you are competing for a slice that is smaller than you think.

Second, the supposedly hot formats are modest. FAQ pages are 3.3%. Reviews are 1.7%. Plenty of GEO guides push FAQ schema as the main lever. The data suggests FAQ pages are a nice supporting actor, not the lead.

The mix changes with the prompt

The split shifts depending on what the user is asking.

Product pages are the top type for all three intents. That is the part I would pin to the wall. Even when someone asks an informational question, ChatGPT is quite often pointing at a page that sells or specifies something.

The branded numbers are worth a look if you handle reputation. When a user asks about you by name, reviews are cited more than three times as often as they are overall, and social posts nearly four times. What people say about you off-site gets pulled in directly.

What to do with this

• Audit your product and pricing pages first. Are specs, prices, availability, and plain-language descriptions stated in text, not hidden in tabs or images?
• For commercial queries, build the comparison and listicle pages that sit next to your product pages. They are the other half of that 70%.
• For informational queries, keep how-to and docs content, but link it to the relevant product page so the path to the commercial page exists.
• Do not expect FAQ pages alone to carry you.

The software category shows the same logic. In Promptwatch's software citation report (Sep 28 to Oct 4, 2026), GitHub, Microsoft's docs, and G2 lead, and documentation and help centers earn the biggest share. Different vertical, same pattern: pages that state facts plainly win.

How big your domain needs to be

This is where the standard advice is most misleading. "Domain authority is key" is repeated everywhere. Promptwatch's August 2026 data, covering the 89.65% of citations whose domain has a known rank, tells a more interesting story. The full domain-rank report has daily charts.

DR 46-75 together is about 46%. DR 61-90 is about 42%. The very top, DR 91-100, got less than the very bottom, DR 0-15. Think about that for a second. The biggest sites on the web are not where most of the citations land.

There is a wrinkle. DR 91-100 share was about 7% in week one of August, around 5% on August 8, and about 3% from August 14, the same day Reddit's share collapsed. Most of the lost share went to DR 61-75 (+1.6 points) and DR 46-60 (+0.9 points). So some of the squeeze on giant domains was a Reddit story. Still, even before the drop, the top band was a small slice.

It also fits the other finding that almost half of ChatGPT responses cite 10 or more unique domains. With that many slots per answer, there is room for mid-sized specialists.

What to do with this

• Stop treating a DR 40 vs DR 70 gap as disqualifying. It is not.
• Aim PR, guest content, and partnerships at DR 61-90 publications. Chasing a placement on a DR 95 giant buys you little by comparison.
• Put your effort into page quality and type before domain-level vanity. Third-party studies find brand mentions correlate more strongly with AI visibility than backlinks do (Ahrefs reported 0.664 vs 0.218 for AI Overviews), which is consistent with a world where being named in mid-tier trusted sites matters more than raw link counts. Treat that as correlation, not a guarantee.

How long a citation lasts

A citation you win in July is not yours in October by default. Promptwatch tracked every page ChatGPT cited on July 6, 2026 for 90 days, and published the content half-life study.

Product pages, reviews, and how-tos reach their half-life at about 74 to 77 days. News is the outlier. Almost a quarter of news pages are never cited again after day 0, and half are gone after 36 days. Docs have a nice property too: 93% were cited again later, versus 77% of news.

Landing pages, docs, comparisons, and listicles only crossed 50% loss in the final week because the window ended, so their real half-life is likely near or beyond 90 days. I would not read the exact numbers as gospel. Read them as relative durability.

What to do with this

• Treat landing pages, docs, and pricing pages as long-term assets. Invest and maintain.
• Keep product specs and prices current. Those pages decay in about 75 days if they go stale.
• Comparisons and listicles are a good investment because they last and they feed commercial prompts.
• Use news and PR for short bursts. Do not expect a press release to keep paying.

Putting the three datasets together

Here is the pattern I take from all of this. ChatGPT cites pages that say something specific and commercial, from mid-authority domains, and it keeps citing durable page types for months.

That last row comes from Promptwatch's social media citations report for Sep 28 to Oct 4, 2026. ChatGPT has nearly stopped citing social platforms while Google AI Overviews still leans on them. A single tactic across engines will not work.

Where third-party advice still holds up

I do not want to throw out everything outside Promptwatch's data. A few claims from other studies are reasonable and compatible:

• Accessibility comes first. A page that is blocked, paywalled, or throwing errors cannot be cited. Digital Applied's meta-analysis of 54 studies, built on Zyppy's scoring, ranks URL accessibility at the top. Check robots.txt and your AI crawler access before anything else.
• Front-loaded answers help. One study reported that 44.2% of ChatGPT citations come from the first 30% of a page. Put the answer early.
• Freshness is real but moderate. The viral "4.3x more cited" statistic traces to no primary study. The better-supported figure is that cited content is about 25.7% fresher than organic top-10 results. Keep pages current, but do not let a refresh calendar replace fixing crawlability.
• Retrieval and citation are separate. One report claims ChatGPT cites only about 15% of the pages it retrieves. Being fetched is not enough.

The ChatGPT fan-out behavior matters here too. On August 8, 2026, ChatGPT began using the site: operator at scale, jumping from about 0.4% to about 17% of fan-out queries (Promptwatch's report). That means it is increasingly running targeted searches against specific domains. Having a clear, well-structured page for each topic on your own site becomes a more direct route to being retrieved.

The screenshot above shows the type of eight-signal framework that is common in agency playbooks. It is a useful checklist. Just weight it using the first-party numbers above rather than treating every signal as equal.

A practical workflow

1. Write down 30 to 50 prompts your buyers actually ask. Mix commercial, informational, and branded. Lock the list.
2. Run them on a schedule across ChatGPT, AI Overviews, AI Mode, and Perplexity. Record which URLs get cited and what content type each is.
3. Sort your own cited and uncited pages by type. If product and landing pages are weak, fix those before writing more blog posts.
4. Check domain tiers of the sites cited beside you. If they are DR 46-75, those are your realistic placement targets.
5. Check crawler access. If AI bots hit errors on your pages, nothing else matters.
6. Set a refresh rule by type: product and pricing monthly, comparisons quarterly, docs when the product changes, news never.
7. Review monthly. The source mix moves, as the Reddit collapse showed.

Tools that help with the tracking

You can do a rough version by hand, but it gets old fast. Tools like Promptwatch track real ChatGPT UI answers, classify citations by content type, and show crawler logs so you can see whether AI bots actually read your pages. It also has a content gap analysis and a Content Agent that can draft and publish pages against gaps. If you want to compare alternatives, the GEO software directory at bestgeosoftware.com lists a wide range.

A few others that come up in this space:

Ahrefs Brand Radar is where the Reddit 16.8% figure came from. Useful for broad domain rankings, though as discussed the source mix has moved.

Profound is an enterprise-grade monitoring option with strong reporting.

Peec AI is a lighter prompt tracker for smaller teams that mainly want mention and citation counts.

For broader ranking comparisons, aisearchtoolrank.com is another directory in this network worth a look.

Common mistakes

• Planning around one number. Citation shares shift month to month. The Reddit drop took days.
• Chasing top-tier domains for placements. The data says DR 61-90 is where the share is.
• Letting product pages rot. They reach half-life in roughly 75 days.
• Betting on news. It loses half its cited pages in 36 days.
• Copying one engine's playbook to all. ChatGPT cites 0.46% social, AI Overviews 12.2%.
• Skipping crawl checks and assuming the content is the problem.

Caveats

These are observational datasets. A page type being cited more often does not prove that switching your page type will get you cited. Classification is automated, and the domain rank data excludes about 10% of citations with unknown rank. The windows are short: one week for content type, one month for domain rank. Revisit them. Promptwatch publishes updates in its data hub, and I would check there before committing a large budget to any single tactic.

Bottom line

If I had to pick three actions from the whole pile: fix your product, landing, and pricing pages so they state plain facts; aim outreach at DR 46-90 sites rather than only giants; and put maintenance effort into evergreen page types while treating news as a short burst. Then measure per engine, on a fixed prompt list, every month.