Your page can be used as a source without you being recommended

A source link in the answer is not the same as being one of the options ChatGPT actually recommends. ChatGPT, Perplexity-style answers, and AI Overviews each reward something different.

Updated 7 October 2026

Someone asks a buying or travel question in Perplexity, and your page turns up as a source.

They ask something similar in ChatGPT, and it names competitors. Your brand, shop, or place doesn’t appear at all.

That is not a broken rewrite. Being used as a source and being recommended are two different outcomes, and a lot of “do GEO” advice treats them as one.

Two wins that look the same from far away

Cited

Your URL is quoted as a source inside the answer. The page did its job for a sourced write-up.

Named

Your product, shop, or place is on the shortlist. Here, competitors are named and you are absent.

Your company page can be cited for one policy or spec while a competitor is the name on the shortlist for “who should I buy from?” Same category, two different jobs. A tourism page can be cited for “is this walking tour running in March?” while a quieter hotel down the street never gets named for “where should I stay near the station?” It also runs the other way: you can be named off the back of your own marketing copy with no independent source behind it, which is what happened in a reported case, a fake deodorant on a narrow ask.

If you are the person asking, this is why a Perplexity answer can feel well sourced while ChatGPT still names a different shop. Either tool can quote a page or name an option, and the log has to record which happened.

The same question looks different in ChatGPT, Perplexity, and Google

ChatGPT, Perplexity, and Google each retrieve differently, so a pattern you see in several of them still isn’t evidence of one shared ranking system.

Hold on to a single realistic question, such as supplements that suit you from a shop in your city, a walking tour instead of the tourist loop, or whether a specific offer is really a deal. Watch what each tool is optimizing for:

Where you asked	What “winning” often looks like
ChatGPT	Your brand, place, or product is one of the options (source links optional or weak)
Perplexity-style answer	Names and links; closest to the GEO paper’s “search, then summarize with citations” setup
Google AI Overview–style block	A summary plus supporting links; availability and shape still depend on query and market

So a well-sourced how-to can feed an Overview or a Perplexity answer while the ChatGPT-style list still leans on whatever its tools retrieved about products or places, which means reviews, Maps, shopping indexes, roundups, and feeds. The sources differ, and so does the result.

Why the mismatch happens

GEO-style edits help when an engine is assembling a sourced answer and needs a page it can quote safely. That effect is real, and it is also limited: the paper’s tests all started from pages that were already in the retrieved set.

Being named on a recommendation shortlist is a different cut of the problem. The assistant still needs a clear identity for you, facts fresh enough to use, and enough corroboration to believe you belong on the list at all. A sharp guide can win the citation while the thing you actually sell, the watch or the hotel or the bottle at this shop, never makes the shortlist. The full pattern is in how AI builds a shortlist.

Schema, feeds, and crawler access each do their own job. Fixing them is no substitute for a citeable page, and a citeable page is no substitute for them.

What to do with that

Decide what a given URL is for.

An authority or explainer page — success means getting cited and leaned on, and GEO-style specificity applies. Treat each section as one question a buyer actually asks, and open with the answer in a few sentences that still stand if the rest of the page is never retrieved.
A product, shop, place, or experience — success means getting named, and described correctly. A citation for some related guide is a bonus, not the same measurement.

Don’t expect a guide-page rewrite on its own to turn into “ChatGPT recommends my product.” When it matters, run the same question in two tools and write down who was named versus who was only linked, before you decide a tactic “works for AI.”

If Perplexity cited you and ChatGPT still named someone else, you haven’t failed at GEO. You’ve found out which of the two games that page was playing. Some people call the citation side AEO (answer engine optimization). It is a useful label for that narrower win. It is not a checklist that also gets ChatGPT to name you.

For how to compare ChatGPT and Perplexity without collapsing it into one score: Those AI visibility tools just count mentions.