Why Low-View YouTube Videos Belong in Your AI Source Strategy
View counts and AI citation presence are different measurement layers, and a company needs to check both.
- Human attention: Public view counts show reach, but view counts alone would not have identified two of the most frequently cited videos in our test.
- AI citation presence: Two affiliate comparison videos with only about 560 and 607 public views were each cited 30 times.
- Commercial interpretation: Review the videos AI answers actually cite before funding new video work or judging video value by reach alone.
What you'll learn#
This article will help you:
- see why audience reach and AI source visibility are different measurement layers;
- recognize that low public view counts did not prevent two videos from appearing repeatedly in this sample's visible citations, so reach metrics alone would not have identified them;
- decide what to examine in the videos AI answers actually cite before investing in more creator content.
Key findings#
- YouTube supplied 131 of the 926 visible citation records we logged across the full experiment — about 14.1% of everything Perplexity cited.
- Citation presence was concentrated. All 131 YouTube records came from just 12 distinct video URLs. Two of those URLs appeared in 30 visible citation records each — together 60 of 131 records, or 45.8% of the visible YouTube layer.
- Those two repeatedly cited videos had only about 560 and 607 public views when we manually checked them. Treat both figures as time-bound observations that will change.
- In the baseline (no-instruction) runs, the visible YouTube share varied substantially by buyer question — from 0% on the selected category question to between 17.2% and 57.8% on the selected product-review and comparison questions.
- Both top videos carried visible affiliate relationships. That is context for reading the source layer, not evidence that either video is inaccurate or that its creator intended anything.
What this means for your company#
Creator reporting and AI source review answer two different questions, and a company needs both.
- Creator reach reporting answers: who attracted human attention — views, reach, engagement, conversions?
- AI source review answers: which specific videos appeared inside the AI answers buyers read when they research and compare products?
Those two lists may not match. In this sample, two videos with low public view counts appeared most often in the visible citations, so reach metrics alone would not have identified their repeated citation presence. A company that ranks its video strategy only by public reach metrics can miss videos that are being cited about its category.
The practical step is to identify the visible video layer first, then assess each cited video on its own terms:
- Accuracy — does the video describe your product correctly?
- Buyer stage — is it cited on a broad category question, a branded review, or a head-to-head comparison?
- Freshness — is it current, or does it describe an older version of the product?
- Ownership or affiliate relationship — is there a visible commercial relationship, such as an affiliate link to a competitor?
- Competitor framing — does the video route the buyer toward another product?
Only after you know whether a commercially important gap exists should you decide what to do about it. The response might involve stronger owned video, better customer or expert evidence, factual corrections, or relevant third-party participation — but this test does not prove that producing any particular video will cause future citations.
How concentrated was the visible YouTube layer?#
Across the full 75-search experiment, YouTube accounted for 131 of 926 visible citation records (14.1%). Those records did not come from 131 videos. They came from 12 distinct URLs — and within those 12, the presence was lopsided.
Two URLs each appeared in 30 visible citation records. Between them, those two videos represented 60 of the 131 YouTube records (45.8%). We watched both.
| Video | Channel | Public views when checked | YouTube citation records | Visible relationship |
|---|---|---|---|---|
| "Fathom vs Fireflies" | Dani's Tutorials | ~560 | 30 | Affiliate link + disclosure |
| "Fireflies vs Fathom" | Software Scope | ~607 | 30 | Affiliate link + disclosure |
Both are affiliate comparison videos: each carries a visible affiliate link and an affiliate disclosure. We report that relationship as an observation. It is commercially relevant context for a company weighing what appears in its category, but it does not establish that either video is inaccurate, and we make no claim about either creator's intent, quality, or independence.
The visible YouTube share also depended heavily on the question. Under the baseline condition — the searches with no source instruction — YouTube supplied 70 of 203 visible citation records (34.5%) overall, but that average hid a wide spread:
| Baseline buyer question | YouTube share of visible citation records |
|---|---|
| "best AI meeting note taker for sales calls" (category) | 0 of 35 (0%) |
| "Fathom AI review" (product review) | 26 of 45 (57.8%) |
| "Fathom vs Fireflies" (comparison) | 19 of 52 (36.5%) |
| "Fireflies AI review" (product review) | 20 of 42 (47.6%) |
| "Fireflies vs Fathom" (comparison) | 5 of 29 (17.2%) |
These percentages describe this selected sample of five baseline questions. They are not an estimate of how often video is cited across Perplexity, this category, or branded queries in general.
What to examine in the visible video layer#
If you run this review for your own category, the useful output is not a popularity ranking. It is a short list of the videos AI answers visibly cite, each checked against the same questions:
- Which videos are cited, and on which buyer questions?
- Are they accurate about your product?
- Who made them, and is there a visible affiliate or competitor relationship?
- Are they current, or do they describe an older release?
- Is there a stronger, more accurate video that could exist for that question instead?
This echoes what we found reviewing the text sources: a visible source layer can carry commercial relationships that are easy to overlook — source instructions expose the evidence layer but cannot rebuild it. Video just makes the relationship easier to see, because the public view count sits right next to the citation.
How we checked this#
We tested one category — AI meeting-note tools — on Perplexity, using five buyer questions (one category question, two product-review questions, and two comparison questions) under five collection conditions: a baseline plus "use official pages," "use reviews," "use Reddit," and "avoid listicles." Each question-condition combination was repeated three times, producing 75 searches in total.
We recorded the 926 visible citation records those searches returned, representing 60 cited domains, and classified each record by domain and source type. We then isolated the YouTube records, counted how many distinct video URLs produced them, and manually opened the two most-cited videos to note their public view counts and any visible affiliate disclosure. The view counts are what we saw at the time we checked; they are time-bound and will drift.
Explore the source data#
You can trace every YouTube citation — and the exact search it came from — in the full source data: 926 visible citation records across 75 searches, representing 60 cited domains, plus rollups by URL and by domain.
FAQ
Can a company rely on YouTube views to identify videos appearing in AI citations?
Not on their own — at least not in this sample. The two videos that appeared most often in the visible citations had only about 560 and 607 public views when we checked, yet each was recorded in 30 citation records — nearly half the visible YouTube layer. Relying on view counts alone would have missed the repeated citation presence of these two low-view videos. This is one Perplexity test; we did not collect view counts for the full candidate set or run a prediction analysis, so it does not establish a general relationship between views and AI citation.
Why should a company track YouTube citations separately from creator reach?
Because they measure different things. Creator reach reporting tells you who attracted human attention; AI source review tells you which videos appeared inside the AI answers buyers actually read. In this test the two videos cited most often had low public view counts, so reviewing only reach would have missed their repeated citation presence.
Does an affiliate relationship mean a cited video is inaccurate?
No. Both top videos carried visible affiliate links and disclosures, which is commercially relevant context for reading the source layer. It does not evaluate accuracy or infer the creator’s intent. An affiliate video can still be correct about a product; the point is to notice the relationship, not to assume dishonesty.
What should a company examine in the cited video layer?
Identify which videos are cited and on which buyer questions, then check each one for accuracy about your product, the buyer stage where it appears, how current it is, any visible affiliate or competitor relationship, and whether a better video could exist for that question. Decide on action only after confirming a commercially important gap exists.
Does this mean every company should produce more YouTube videos?
No. This test shows which videos appeared in one set of visible citations; it does not prove that publishing a new video will cause AI citations. The first step is to see which videos are already cited about your category and whether they represent you accurately — production decisions follow from that, not from a general rule to make more videos.