Learn how to improve your product’s position in ChatGPT answers by analysing cited pages, source order, page types and answer stability.
The false premise behind “how to rank in ChatGPT answers” is that ChatGPT gives your page a search position to improve. The practical target is usually your product’s position on the source page the answer uses, and the order in which the assistant names products.
You are not trying to move a URL up a results page. You are trying to get your product into a relevant roundup or comparison, then make its fit clear enough to earn a strong position there.
An assistant can compress a source page into a buying suggestion. If the source page names Product A before Product B, the assistant has a clear route to naming Product A first. If the assistant follows source order, the first-named product is the one buyers see as the default pick. The ledger does not tell us why buyers weight it that way, only that order maps through. A citation confirms that the page entered the answer, but it does not show which product received the strongest position.
The label around your name matters as much as the name itself.
The pattern in the CiteGraph research is useful here: 45,309 recorded citations came from 7,260 AI answers. Listicles and roundups supplied 67.5% of citations, while competitor product pages supplied 24.4%. Together, those page types account for 91.9% of citations in the sample, which makes source-page order a sensible place to start.
Start with the page that supplies the recommendation, not your own position report.
A source page can put your product first, place it lower down, describe it as an alternative, or include it in a comparison where the surrounding text determines the meaning. These are different commercial positions even when the cited URL stays the same. You need to measure citation presence and answer position separately, because one can improve while the other stays unchanged.
Source order is the part you can inspect directly. Find the pages that answer the buyer questions you care about, record where your product appears, and read the words around that position. The page may call one product best for a particular use case, another easiest to set up, and another better suited to a smaller team. Clear distinctions give an assistant usable material when it turns a page into an answer.
The URL alone cannot tell you who won the recommendation.
A useful review asks six things:
This is the check that citation tracking alone will not do for you: it stops at presence, not position.
For a worked example, take the page-type figures in the CiteGraph research sample. Start from the total: 45,309 citations. Listicles and roundups took 30,585 of them, so divide 30,585 by 45,309 and you get 67.5%. Competitor pages took another 11,052, or 24.4%. Add those two shares and you land on the 91.9% figure cited above. The remaining page types account for the balance.
The arithmetic shows why page choice matters, but it does not prove that every answer copies every source page exactly.
Apply the same review to a cited roundup. If a competitor appears first in the section matching the buyer’s need and your product appears lower down, the roundup has entered the citation path but the competitor owns the strongest source position. You then need to understand whether the gap comes from the page’s order, its category labels, its evidence, or your product’s fit for that use case.
Prioritise pages that already influence the questions you care about. Look for roundups, comparison pages and competitor product pages that a buyer could plausibly use to choose between products. Then map each page to a specific buyer question, because a page can use one order for small teams and another for regulated businesses. A page has little to work with if your positioning is “best for everyone”, so give it a specific job to assign you instead.
Match the position to the buyer’s reason for asking.
Use your own site as supporting evidence. Make the product’s category, audience, strengths, limits and alternatives easy to verify, with key facts in normal page text and consistent product names. The product’s own site accounts for 1.5% of recorded citations, or 663 citations, in the sample, so it can support the preferred description without being the main route to a roundup position.
Your site should make the preferred description easy to copy.
The practical workflow is short: find the cited page, check where your product appears, identify the buyer question it answers, compare your description with the product above you, supply evidence for the position you want, then recheck the source page and answer. The free AI visibility checker can speed the citation-review step.
A third-party page decides the visible order, so your work is partly editorial rather than purely technical.
Competitor product pages deserve the same inspection. They may already sit inside the observed answer set and contain comparisons your own site would not publish about itself. Read the description, order and alternatives, then decide whether the gap is a content problem, a distribution problem or a product-fit problem.
A strong source page does not need to praise you without qualification. It needs to explain the fit clearly enough that the order makes sense.
Record the cited URL, the product order on that URL, the product order in the answer and whether the source page explains the recommendation. Keep the buyer question and prompt fixed when comparing results, and save the answer text as well as the cited pages. This gives you a path from a page change to an answer change instead of one visibility score that hides the failure point.
Track position, not just presence.
The wider sample covers 60 scans of 41 sites in 44 categories and 442 distinct buyer questions. The engines cited 7,624 distinct domains and 16,506 distinct pages. In CiteGraph’s own project, 34 of 40 heavily cited domains from the first scan were still cited 17 days later, while 41 of 179 exact pages appeared again; counting every cited domain, 69 of 102 returned.
Domains were more stable than exact URLs in that comparison.
A domain can remain visible while the page carrying the recommendation changes. If you track domain mentions alone, you can miss a new roundup, a changed order or a page that now gives a competitor the first position. The methodology page sets out how CiteGraph measures these observations.
We measure samples, print the margins and do not know what we have not measured. A product can hold the best position on a cited roundup and still be absent from an answer outside this sample.
No. The more direct target is often your product’s order on the roundup or comparison page that ChatGPT cites. A page can be cited while another product holds the stronger position on that page.
Save the answer and its cited pages for a fixed buyer question. Check your product’s order in the answer, then inspect whether the cited source page places it first for the same use case.
Only when it fills a clear evidence gap. Existing roundups, comparisons and competitor product pages may offer a more direct route because they already appear in the observed citation pattern.
They can change when the answer or cited page changes. In CiteGraph’s own project, heavily cited domains were more persistent than exact pages across the 17-day comparison.
Cite this: CiteGraph, “How to rank in ChatGPT answers, not search results”, 11 Sept 2026, https://www.citegraph.app/blog/how-to-rank-in-chatgpt-answers