How We Choose PR Targets Using ChatGPT Domain Rank Data
Promptwatch's August 2026 report groups ChatGPT citations by domain rating. Here is how we use it to decide where client PR and guest content should aim.
Sites with a domain rating of 91 to 100 earned 4.4% of ChatGPT citations in August 2026. Sites rated 61 to 75 earned 23.4%. That contrast, from Promptwatch's report on citation share by domain rank, changes how we brief clients on PR. The report was published on September 25, 2026, and covers August 1 to August 31, 2026.
The usual client request is a placement on the biggest publication we can get. We still pitch those when there is a real story. But we no longer treat them as the default target for AI search purposes, and this is the reasoning.
What the report shows
The report covers 89.65% of ChatGPT citations, meaning the domains with a known rank. The other 10.35% are excluded, so every share below is a share of the ranked portion, not of everything.
By group, the monthly averages are: DR 61 to 75 at 23.4%, DR 46 to 60 at 22.6%, DR 76 to 90 at 18.9%, DR 0 to 15 at 6.3% and DR 91 to 100 at 4.4%. Put together, DR 46 to 75 is 46%, almost half. DR 61 to 90 held about 42% all month. Low-rated domains matter too: DR 0 to 30 is about 14%, and DR 0 to 45 is almost a third.
The top tier moved during the month. DR 91 to 100 averaged 7% in the first week, about 5% on August 8, and about 3% from August 14. That is the same day Reddit's share collapsed, which the report notes. The largest gains went to DR 61 to 75 (up 1.6 points) and DR 46 to 60 (up 0.9).
What it does not tell us
It is one month, one engine, one vendor's prompts. Domain rating is a third-party metric, and the report treats it as a way to group domains, not as something ChatGPT is known to use. The data does not show that a DR 70 site is cited because of its rating. It shows where citations landed. The fall in the top tier is also tangled up with Reddit's August drop, which we cannot separate out from this data. If we pitched only DR 61 to 90 sites on the strength of one month, we would be overreading it.
How it changes our target lists
The report's suggestions are specific, and we attribute them to it: don't wait for a top-tier DR, put less weight on placements on the biggest sites, aim PR and guest content at DR 61 to 90, and benchmark against DR 46 to 75. Here is what we do with that.
- Build target lists in tiers. The primary list is publications in the DR 61 to 90 band that actually cover the client's topic. We still keep a short list of top-tier outlets for stories that earn them.
- Put relevance first. A DR 75 site that never writes about the client's category is a poor target. A DR 55 trade publication that readers in the field use is often better. The rank band narrows the list. It does not choose the target.
- Drop the DR threshold as a gate. Many clients have been told never to accept a link below some number. For this purpose, the data suggests the cutoff is not helping. Nearly a third of citations in the report went to domains below DR 45.
- Keep the human outcomes in view. PR also has audiences, referral traffic and reputation to think about. AI citation is one reason to choose a placement, not the only one.
We do not promise that a placement leads to a citation. A good story on a well-chosen site is a reasonable bet, and we treat it as one.
A list we can build without tools
Start with the client's topics and the sites that cover them. Look up each site's rating in whichever SEO tool you already use. Sort them into the bands above. Then read the last ten articles on each site to see whether they ever mention a company like the client's. Anything that never covers the category falls off the list, no matter its rating. What is left is a target list you can defend to the client and the PR team.
The measurement step
The step that is hard to do by hand is seeing which third-party domains ChatGPT cites for the client's own prompts. We do that in Promptwatch, the platform we run client programs on. Domain-level citations show which sites appear in answers to the tracked prompts, and page-level citations show the specific URLs. We compare that list against the DR bands from the report and against the client's competitors.
Offsite mentions are the most useful view for PR planning. They show when a brand is named on third-party pages without a link, and where. After a placement lands, we add the URL to Page Tracker and watch whether any tracked response cites it. We read the result cautiously. A placement that does not get cited in the first weeks is not a failure, and one that does is not proof of anything beyond that window.
Citation Trends adds time. It shows first-cited date, ramp-up, peak rate and decay for each page, so we can see whether a placement is climbing or has already peaked. That feeds the next round of pitches.
What we tell clients
The report suggests the middle of the rank range does a lot of the work in ChatGPT's citations. So we aim our PR effort at relevant, well-edited sites in that range and keep the very top for stories that deserve it. Then we measure against the client's real prompts and adjust. A month of data from someone else's prompt set is the opening hypothesis, and the client's own tracked citations are the evidence that replaces it.