Video Editing Market Size: We Analysed 2,417 Keywords (2026 Study)

The video editing market is far smaller than the video category that contains it. We analysed 2,417 keywords and 318 of our own published articles, and found that only 2.5% of US video search demand comes from someone trying to hire an editor. The remaining 97.5% is people looking for software or people looking to film something, which are different businesses with different buyers.

This is a study of our own data, published in full including the parts that make our previous strategy look wrong. We are a video editing agency, so we had a commercial reason to hope the number was larger. It is not.

Key findings

1. Buyer-intent demand is 2.5% of the category. 368 keywords, 58,020 US searches a month, against 1.9 million a month for software and DIY tools.
2. Ranking on page one is close to meaningless at this scale. Across 109 of our pages that rank on page one for at least one query, the median page earns 3 impressions in 90 days.
3. Keyword format predicts performance better than topic. Comparison pages earn a median of 152 impressions. Industry pages earn 6, and we built 49 of them.
4. Question-shaped queries produced 643 impressions and zero clicks. Not a low click-through rate. Zero.

Method

Keyword volumes come from DataForSEO's Google Ads endpoint, US only, pulled on 6 August 2026 across four datasets: seed expansion from our own terms, related-keyword expansion, and a full crawl of a direct competitor's ranking keywords. Total cost of the pull was under a dollar, which is worth stating because the finding did not require an expensive tool.

We then classified each keyword by what the searcher plainly wants: software, filming, or hiring an editing service. 1,558 of the 2,417 keywords fell cleanly into one of the three. The remaining 859 were ambiguous, brand-specific or unrelated, and we left them out rather than forcing them into a bucket. All percentages below are shares of the 2,312,060 monthly searches represented by the classified set.

Performance data comes from Google Search Console for pixel8production.com, covering 11 May to 9 August 2026. Search Console's query dimension omits anonymised queries, so query-level figures describe 8,134 of our 20,329 total impressions. Site-level totals use the unfiltered numbers.

Finding 1: the market is 2.5% of what it appears to be

What the searcher wants	Keywords	US searches/mo	Share
Software and DIY tools	1,017	1,903,350	82.3%
Shooting and production	173	350,690	15.2%
Hiring an editing service	368	58,020	2.5%

The video industry's headline numbers describe the first row. Someone searching "video editing programs" or "canva video editor" is not a customer for an editing service, and a competitor ranking for those terms is harvesting traffic that will never convert into a retainer. When we crawled one competitor's keyword profile, their two largest terms were "video editing programs" at 301,000 searches a month and "canva video editor" at 40,500. That looks like dominance and functions as noise.

Even inside the 2.5%, the segment is not cleanly addressable. The single largest term is "video production service" and its plural, at 6,600 searches a month, and both imply filming. We do not film. The genuinely servable head terms are "video editing service" and its plural at 2,400 a month, and "hire a video editor" at 1,600.

So the realistic ceiling for a specialist editing business, in US search, is roughly 58,000 monthly searches, of which the head terms account for a few thousand. Ranking first for everything would still be a modest number of visitors before click-through and conversion.

Finding 2: page-one rankings that earn nothing

We published 318 articles. Here is what they produce.

Metric	90 days
Clicks	117
Impressions	20,329
Click-through rate	0.58%
Average position	11.9
Pages with at least one impression	293

The natural diagnosis is that the writing or the rankings are bad. Neither holds up.

109 pages rank on page one for at least one query. 97 of them, or 89%, earn under 50 impressions in the full 90 days. The median page-one page earns 3 impressions. At the query level it is starker: of 253 queries where we rank in the top ten, the median earns 2 impressions in 90 days, and 71% earn five or fewer.

We have pages sitting at position 2.0 with a single impression. That is not a ranking problem, and no amount of rewriting fixes it. Those pages were built on keywords whose volume nobody checked, and they rank first for questions almost nobody asks.

This is the mechanism behind the headline number. A corpus can be well written, correctly optimised, and genuinely ranking, and still produce almost nothing, because the demand underneath it was never measured.

Finding 3: format predicts performance better than topic

Grouping our pages by the shape of the keyword rather than its subject produced the clearest predictor in the dataset.

Keyword class	Pages	Median impressions (90d)
[brand] alternatives	10	152
best or top [X]	10	47
Cost and pricing	12	26
[X] vs [Y]	32	14
[thing] service	65	12
video editing for [industry]	49	6

A comparison page outperforms an industry page by roughly 25 to 1. The uncomfortable part is the page count: our single largest content investment, 49 industry pages, went into the worst-performing class in the set. They were easy to produce and templated well, which is exactly why there were so many of them, and templated production is a good way to manufacture pages nobody searches for.

The pattern is not mysterious. Someone typing "descript alternatives" is comparing purchases. Someone typing "video editing for dentists" is usually a dentist idly wondering, if they exist at all.

Finding 4: the clicks that never existed

We split queries by whether they are phrased as questions, as a rough proxy for the queries that AI answer surfaces tend to resolve directly.

Query type	Queries	Impressions	Clicks	CTR
Question-shaped	96	643	0	0.00%
Everything else	670	7,491	24	0.32%

Zero clicks from 643 impressions. Some of those queries sit at strong positions. The impression is real, the position is real, and the visit does not happen, because the answer was delivered on the results page and the user had no reason to continue.

Position bucket data says the same thing from another angle:

Position	Queries	Impressions	Clicks	CTR
1 to 3	29	318	11	3.46%
4 to 10	221	3,605	7	0.19%
11 to 20	224	2,235	5	0.22%
21+	268	1,507	0	0.00%

Positions four to ten should convert at a few percent. Ours convert at 0.19%. That gap is not explained by titles or meta descriptions alone. It is what a results page looks like when the top of it answers the question before anyone scrolls.

For a business, the implication is awkward but clear: on question-shaped queries, being ranked is not the goal, because the click does not exist to be won. Being cited is the goal, and it is measured differently.

What we changed

We stopped scaling article volume. Publishing more pages into a 58,000-search market produces more pages, not more customers.

We stopped building video editing for [industry] pages, having built 49 of them.

We began entering the adjacent production market honestly. We do not film, so we cannot serve the 350,690 monthly searches for shooting directly, but we can be the editing partner inside that work, and we say so plainly rather than implying a capability we lack. Our guides to individual city markets now rank production companies alongside us and state which jobs we are the wrong choice for.

We moved effort off-page. When the addressable search market is this small, links, citations and referrals matter more than the next article.

And we published our pricing, which is $900 for a one-time trial and then $2,000 to $3,200 a month depending on volume. Not a finding from this study, but related: in a small market, disqualifying the wrong buyer early is worth more than another visitor.

Limitations

This is one site's data in one niche, and the corpus was built by one team with a particular set of habits. The keyword classification is our judgement, not a standard, and a different analyst would sort some terms differently. Volume figures are Google's estimates and are known to be imprecise at the tail.

Two specific cautions. First, volume data should not override observed data. One of our terms reports 10 to 20 searches a month in Google Ads and produced 1,027 real impressions in Search Console over 90 days. Where the two disagree, trust the one measured on your own site.

Second, the question-shaped split is a proxy. We cannot see AI answer surfaces in Search Console, so we are inferring their effect from a pattern in the click data rather than observing it directly.

Using this data

The figures here are free to cite with attribution to Pixel8 Production and a link to this page. If you want the segment definitions or a breakdown we have not published, ask and we will share it.

We publish this partly because it corrects things we previously believed and acted on. The market did not turn out to be the size our content plan assumed, and 49 pages exist as evidence of that. Anyone selling video services on the strength of the category's headline numbers is working from the wrong denominator, and it is cheaper to learn that from someone else's data than from three hundred of your own articles.

If you are choosing between models rather than measuring markets, our comparison of a video editing subscription versus a freelancer covers where each one stops working.