Most Perplexity Citation Stats Are a Year Out of Date

Most published figures on Perplexity citations report something in the range of 6 to 8 sources per answer. The Orbit Media study from August 2026 measured 19.2 numbered source citations per answer.

That gap -- 6 versus 19 -- is not random noise. It reflects a platform that has substantially expanded its citation display over 18 months and a research ecosystem where most published figures are still catching up. If you're reading AEO advice that references Perplexity's citation behavior, the first question to ask is when the study was conducted.

Why Published Citation Counts Disagree by a Factor of Three

In our 2026-10-23 cross-study calibration of per-platform citation counts (`platform-citation-behaviors.md`, Scout session 177), we tracked down five separate citation count studies for Perplexity and mapped why they diverge.

| Source | Data period | Citations/answer | |---|---|---| | xfunnel.ai (40K answers) | ~late 2024 | 6.61 | | Texta analysis | Jan-Mar 2026 | 4-8 | | Orbit Media (13,184 citations) | Aug 2026 | 19.2 | | LLMPulse ranking table | 90-day window (Jul 2026) | 2.57 | | LLMPulse glossary | undated | 21.87 |

The LLMPulse discrepancy -- 2.57 on one page and 21.87 on another page of the same site -- is a methodology mismatch between pages. Neither figure is reliable without knowing the source.

The more meaningful comparison is xfunnel.ai (6.61, late 2024) versus Orbit Media (19.2, August 2026). Two causes explain most of this gap.

The first is timing. Perplexity genuinely expanded its source display across 2025 and into 2026. xfunnel.ai measured a different product. The second is what each study counted. xfunnel.ai tracked unique outbound links embedded in the response text. Orbit Media counted numbered source cards displayed in the Perplexity UI. Perplexity typically shows more source cards in the sidebar than it embeds as inline links in the answer text -- which means both figures can be accurate for their respective measurements while still describing very different numbers.

For planning purposes, use Orbit Media (19.2) as the current working estimate. Any study reporting 6-7 for Perplexity is referencing a platform that looked substantially different a year and a half ago.

What Went Into Those Additional Slots

More source slots means more citation opportunities -- but only if you understand which sources are actually being retrieved. The assumption most people bring to this is that Reddit is a central part of the answer. It is not.

In our 2026-05-13 empirical analysis of Perplexity source URLs from a real Sourcepull audit (`platform-citation-behaviors.md`, session 16), we analyzed 102 source URLs captured across 12 Perplexity query records for a single business.

Zero were from Reddit.

That audit covered a B2B software category, where the 4 category-level queries produced the densest citation data. The domains appearing most consistently for those queries were niche comparison sites and category pages on industry directories: Profound, Meltwater, Conductor, G2's category overview page (not individual product pages), Stackmatix, and Hubspot.

The G2 category page finding is worth pausing on. When someone asks Perplexity "what are the best AEO audit tools," Perplexity cites the G2 category page -- the overview that lists and compares products -- not individual product listings. That page appeared in 3 of 4 B-queries in the audit. Appearing on the G2 category page, even at a low rank, is what gets a business into Perplexity's citation pool for category queries in that vertical.

Every business category has an equivalent. For a plumber in a major market, Perplexity is pulling from a specific set of 3-5 sources when someone asks who to call. Those sources are not the same 25 directories on every generic AEO checklist. They're the specific comparison articles, local service comparison sites, and vertical directories that Perplexity's live retrieval is currently treating as authoritative for that category in that city.

The Reddit Timeline

The conventional narrative in early 2026 -- that Reddit presence was essential for Perplexity visibility -- was outdated before most AEO guides were published.

In our 2026-04-23 investigation of Reddit as an AI citation signal (`reddit-as-aeo-signal.md`, session 3, updated through Q2 2026), we traced the full timeline. At peak -- mid-2025 -- approximately 46.5-47% of Perplexity responses included at least one Reddit citation. This was the measurement driving the "Reddit is essential for Perplexity" claim.

In October 2025, Reddit filed a federal lawsuit against Perplexity for data scraping (SDNY, filed October 22, 2025), alleging use of proxies and IP rotation to circumvent their access restrictions. Within weeks of the lawsuit filing, Perplexity's Reddit citation share dropped approximately 86%.

The Tinuiti Q2 2026 tracking showed the result: approximately 7% of Perplexity responses included any Reddit citation by April 2026. Down from 47% the prior year.

The sources that absorbed Reddit's share in the citation pool were niche expert directories and vertical comparison sites -- which is consistent with what the session 16 source_url audit data showed directly. The sites filling the gap were the industry-specific sources Perplexity's retrieval already preferred for structured comparison queries. Reddit's collapse made those sites' dominance more visible, not larger.

The Retrieval Architecture Explains the Source Selection

Perplexity is a real-time retrieval engine. Every query triggers a live web search. This architecture has two implications most AEO advice undersells.

The first is that Perplexity's citation pool is dynamic in a way that ChatGPT's is not. ChatGPT's citation behavior for a given category reflects training data that's months old and partnership data (Yelp, Foursquare, Thumbtack) that's updated on a slower cadence. A change to your ChatGPT visibility takes weeks to months to register. A change to your Perplexity presence -- a new listing on a comparison site Perplexity is already pulling for your category -- can show up in citations within days.

The second implication is query-vocabulary sensitivity. Perplexity's retrieval executes close to the original query phrasing. Our 2026-06-25 analysis of retrieval architecture differences (`platform-citation-behaviors.md`, session 58) documented that Perplexity keeps approximately 88% of the original prompt words and runs roughly one retrieval query per user prompt. ChatGPT generates 91% unique sub-queries from the same prompt, fanning out into related angles.

This means Perplexity is more literal. A business listed as "water heater installation" is retrieved for queries that use exactly that language. A business listed only as "plumbing services" is less likely to appear when someone asks specifically about water heaters. Vocabulary alignment with the specific services described in comparison sites and directories is more valuable for Perplexity than content breadth is for ChatGPT.

The Practical Implication for Fix Sequences

For a business that scores weak on Perplexity category queries and better on brand queries, the standard generic directory checklist is not the problem. Standard directories are probably covered. The gap is absence from the 3-5 niche sources Perplexity is currently pulling for that category.

Identifying those sources requires capturing what Perplexity actually cites for the relevant category queries -- not what AEO advice suggests Perplexity should cite. The source_url data from audit records, when analyzed per query, shows directly which domains are functioning as Perplexity's category authority for a given vertical and location. That list is the fix plan.

Our session 16 analysis produced that list for one category and found it was almost entirely distinct from what a standard AEO fix plan would recommend. The comparison article sites driving category citations weren't on any AEO checklist. They were discoverable only by looking at what Perplexity was actually retrieving.

Signal Check at sourcepull.ca includes Perplexity citation rates for both brand and category queries. For businesses with weak category scores despite established directory presence, the full audit report surfaces the specific Perplexity source domains driving that gap.