• AI visibility

Guide · AI search

How do AI video companies win customers through AI search?

By being one of the tools an AI assistant names when a creator, marketer or training team asks what to use, and by making the free plan, the price and the trust questions easy to check once they arrive. AI video is a crowded, fast-moving category where buyers try several tools and switch cheaply, so being on the shortlist matters more than being first in a list of links. No published data yet shows how much of any AI video company’s revenue starts in an AI answer, so treat this as a channel to measure, not a proven one.

By Underneath Team Guide Updated

The short version

  1. The category is large and concentrated at the top. Synthesia (opens in a new tab) reports $150 million in annual recurring revenue (ARR) and 60,000 business customers, including 90% of the Fortune 100; HeyGen (opens in a new tab) says it passed $200 million in ARR in June 2026, doubling in eight months.
  2. Use is now mainstream: 63% of video marketers in Wyzowl’s 2026 survey (opens in a new tab) had used AI video tools, up from 51% a year earlier, and 52% of learning and development (L&D) teams in Synthesia’s 2026 report use AI for video creation.
  3. Buyers shop around. In Adobe’s survey (opens in a new tab) of 16,000 creators, 60% used more than one generative AI tool in three months, and 58% find new tools through personal research.
  4. The money path is short and self-serve: Synthesia’s pricing (opens in a new tab) runs from a free plan to $29 and $89 a month before enterprise contracts, and Runway’s (opens in a new tab) paid plans start at $12 a month billed annually.
  5. An assistant can know a tool’s name without ever recommending it. In a test of 112 Product Hunt startups, Sharma (opens in a new tab) found ChatGPT recognized 99.4% by name but surfaced only 3.32% in “what are the best tools” questions.

Who pays for AI video tools, and how much is each kind of customer worth?

Three groups buy it: creators, marketing teams and enterprise training or communications teams, each worth very different amounts.

Creators and solo businesses buy on a card. Adobe’s Creators’ Toolkit Report, run with The Harris Poll in September 2025, found 86% of creators actively use creative generative AI, and 52% use it to generate new assets such as images and video. They are price-sensitive: the top barriers to adopting a tool were high cost (38%), unreliable output quality (34%) and uncertainty about how the AI model was trained (28%).

Marketing teams buy for volume. Wyzowl, which surveyed 266 respondents in late 2025, found 91% of businesses use video as a marketing tool, and 69% of video marketers have made social media videos, the most common use. Explainers (68%), product demos (39%) and sales videos (37%) follow.

Enterprise training, HR and communications teams buy for scale and languages. Synthesia’s 2026 L&D report, based on 421 responses, found 87% of respondents already using AI, with voice generation (63%) and video creation (52%) among the main uses. HeyGen says 85% of the Fortune 100 use it “for training, sales enablement, and localization.” Voice tools sold to the same teams are covered in how AI voice companies win these buyers.

What a customer is worth follows the same split. A creator may pay tens of dollars a month. A company that standardizes on one platform for training in dozens of languages signs an enterprise contract and then grows it. Synthesia said it reached $100 million in ARR (opens in a new tab) through new customers “across all pricing plans” and that more than 70% of the Fortune 100 were customers by April 2025, up from 40% two years earlier. Tech.eu reported that the company credited existing customers spending more (opens in a new tab) as one of its growth drivers.

At what point do AI assistants enter the search for a video tool?

At the research step, where buyers look for a tool, and increasingly inside the assistant itself.

Adobe found creators scout and test new tools through personal research (58%), social media trends (57%) and recommendations from other creators (41%). The survey did not ask which share of that research happens in ChatGPT or Gemini, so we cannot give an AI-specific figure for AI video buyers. Outside video, G2’s 2026 survey of software purchasing (opens in a new tab) found 51% of B2B software buyers now start research with an AI chatbot more often than with Google. Enterprise video platforms are bought the same way, so a reasonable expectation is that their buyers behave similarly.

Two developments make AI assistants more than a research tool in this category:

  • The assistant is becoming a place to make the video. HeyGen documents a ChatGPT app (opens in a new tab) that lets users describe a video in ChatGPT and have HeyGen’s agent produce it without leaving the chat. For a buyer, the first trial may now start inside the assistant.
  • The audience is spread across assistants. Similarweb’s 2026 data (opens in a new tab) shows ChatGPT’s share of generative AI website visits falling from about 76% in June 2025 to roughly 53% by May 2026, while Gemini rose to around 27–28%. A video company visible only in ChatGPT reaches a shrinking share of researchers.

Google’s own AI answers matter too. In our study of 1,248 US searches, B2B software and technology keywords showed an AI Overview on 96.0% of searches, and in our YouTube study the AI Overview cited a YouTube video on 91.0% of B2B software searches. These were general software searches, not AI video ones, but they suggest that tutorial and review videos about your tool are part of what Google’s answers draw on.

Which questions do AI video buyers ask?

Mostly fit questions: which tool for a use, which alternative to a known brand, and the cost per minute.

We wrote these example prompts ourselves to show how creators, marketers and training teams tend to frame a video-tool search; they are not observed data:

  • Use case: “Best AI avatar tool for compliance training videos in 20 languages.”
  • Alternatives: “Cheaper alternatives to Synthesia for a five-person marketing team.”
  • Head-to-head: “HeyGen vs Synthesia for personalized sales videos.”
  • Model choice: “Which AI video generator makes the most realistic product b-roll: Runway, Kling or Veo?”
  • Rights and risk: “Can we use an AI avatar of our CEO in ads that run in the EU?”
  • Price: “How many minutes of video do I get on HeyGen’s cheapest paid plan?”

Vendors already compete on these questions in public. HeyGen’s own blog lists a post titled “15 Best Synthesia Alternatives in 2026, Ranked by Real Render Tests.” Price questions are where answers slip most easily. Credit-based plans are hard to summarize, and in our pricing study, only 61.9% of the plan prices four assistants quoted for 45 software products were fully faithful to the vendor’s pricing page.

How does an assistant’s mention become a paid video plan?

Through a free plan: the assistant names you, the buyer signs up, the credits run out, and the team upgrades.

Named. The assistant puts your tool on a short list for the buyer’s use case. Creators rarely stop at one: 60% of Adobe’s respondents used more than one generative AI tool in the past three months.

Free plan or trial. The cost of trying is close to zero. Synthesia’s pricing page lists a $0 plan with 10 minutes of video a month; Runway offers 125 one-time credits to explore its tools.

Paid self-serve. Synthesia’s paid plans are $29 and $89 a month, and Runway’s are $12, $28 and $76 a month billed annually. Usage is metered in credits or minutes, so revenue grows with use.

Team and enterprise. Training and communications teams move to contracts for security, seats and languages. Synthesia’s report found budget approvals (44%) and procurement processes (19%) add friction at this step, which is where a buyer goes back to research and asks follow-up questions about security and compliance. AI coding tools follow a similar climb; see how coding tools grow from one developer to a team.

The click itself may not show up as AI traffic. Similarweb reports that roughly six in ten ChatGPT referrals now land on a homepage, and that traffic filed as “direct” may increasingly be AI-driven discovery. HeyGen credits much of its growth to a product that spreads “largely through word of mouth.” We infer that, for video companies, part of what looks like direct or word-of-mouth signups may begin in an AI answer, and only a signup survey question will show it.

Why would an assistant suggest one video generator over another?

Platforms document little; studies point to independent coverage, community presence and fresh pages; buyers add consent and rights.

Documented by the platform. No major assistant publishes how it picks the tools it recommends for a creative use case. What is documented is the rule set around the output: under Article 50 of the EU AI Act (opens in a new tab), which applies from 2 August 2026, providers of systems that generate video must mark output as artificially generated, and deployers who publish a deepfake must disclose it. Lewis Silkin notes (opens in a new tab) that this is likely to capture a lot of AI-generated advertising.

What research has found. On US software ranking questions, AI search took 72.7% of its sources from independent “earned” sites, against 45.4% for Google, in work by Chen and colleagues (opens in a new tab). Sharma’s Product Hunt study, whose sample leaned toward developer, productivity and AI products, found that startups whose own sites scored higher on GEO-style content checks were no more likely to be surfaced, while the number of other sites linking to them and community discussion did predict Perplexity visibility. Freshness matters in a category that ships new models every few months: in our freshness study, four assistants cited pages first published about half as long ago as Google’s top 10 for the same questions (a ratio of 0.50).

Trust factors specific to AI video. Buyers ask about things other software categories rarely face:

  • Consent and likeness. Synthesia describes a framework that requires explicit consent for all avatar likenesses. Whether an avatar of a real person can be created and published is a buying question in itself.
  • Training data and rights. In Adobe’s survey, 69% of creators were concerned about their content being used to train AI without permission. Image tools answer similar questions; see how image generators handle commercial rights.
  • Security and governance. Synthesia’s L&D respondents named security (58%), accuracy (52%) and legal constraints (41%) as major obstacles. Synthesia announced in September 2024 that it was the first AI video company (opens in a new tab) to achieve ISO/IEC 42001 certification for AI management.
  • Reviews. HeyGen reports that it won 281 badges in G2’s Summer 2026 reports. A reasonable expectation is that review profiles feed both buyers and the answers they read.

What does a video tool lose when assistants skip it?

Usually the free signup, and with it the chance to grow into a paid or enterprise account.

The category moves fast and punishes a missing name. In a16z’s March 2026 ranking (opens in a new tab) of consumer AI products, video generation “saw the most movement,” with Kling AI, Hailuo and Pixverse building traction and Google’s Veo 3 driving traffic to Google Labs. The same report notes that where Google and OpenAI focus their creative efforts, including video, standalone traffic compresses. When the model makers’ own tools sit inside the assistant, an independent video company that the assistant does not name has little else to fall back on. AI research tools face the same pressure from built-in research modes; see how research tools earn trust.

Low switching costs cut both ways. A creator who tries three tools in a month can find you from a single answer, and leave for a rival named in the next one. In an enterprise deal, the cost is larger: once a training team standardizes on one platform and builds its library there, the account tends to stay, so missing the first shortlist can mean missing years of seat and usage revenue. That last point is our inference from how these contracts are sold; no study has measured it for AI video.

What can GEO do for an AI video platform, and what can’t it?

It makes your tool easy for assistants to describe, compare and trust; it cannot promise a placement in any answer.

For a video platform, generative engine optimization (GEO) work usually falls into seven areas:

  1. A clear category and use case. Say plainly whether you are an avatar platform, a generative model, an editor or a localization tool, and for whom. Assistants answer use-case questions, so “training videos in 140 languages” is easier to match than “the future of video.”
  2. Independent reviews and coverage. Creator reviews, tutorials on YouTube, practitioner newsletters, L&D and marketing publications, and review platforms. Those outside voices are what AI search tends to draw on when it names video tools. Our article on the pages AI engines cite when naming brands covers which ones matter.
  3. Fair comparison pages. Publish honest comparison and alternatives pages with real test conditions, the way buyers already ask. Video vendors weighing a “HeyGen vs Synthesia” style page can check the citation evidence in our summary on comparison pages.
  4. A pricing page an assistant can quote. State what each plan includes in minutes as well as credits, keep one current page, and retire old figures that assistants can still find.
  5. Trust pages in plain words. Consent rules for avatars, how models were trained, commercial-use rights, EU labeling support and security certifications, on public pages rather than behind a sales call.
  6. Fresh pages for each release. New models and features change the answer to “which tool is best for this” every few months. Dated, specific release pages and updated benchmarks give assistants something current to cite. Our article on why ChatGPT misses new products explains the gap launches face.
  7. Measurement across assistants. Track a fixed set of use-case, alternatives and price questions in ChatGPT, Gemini, Perplexity, Copilot and Google, and add “how did you hear about us?” to signup.

What is still unknown about AI search and video tool sales?

Nobody knows how much AI video revenue starts in an AI answer; no company in the category has published it.

The revenue figures here are company-reported. The adoption surveys are small (Wyzowl’s 266 respondents) or run by vendors with an interest in the result (Synthesia’s L&D report, Adobe’s creator survey). The best evidence on how AI search picks software comes from general software questions and from Product Hunt startups, not from AI video specifically. No published study yet follows an AI recommendation through to a paid video plan or an enterprise contract, and none shows how assistants handle questions about consent or deepfake rules. Whether visibility work pays off at all is the subject of our article on business results.

How should an AI video company find the answers that cost it signups and upgrades?

Check which AI answers name you for the use cases that bring in your paying customers.

A practical first step is an audit of your use-case, alternatives, model-choice and pricing questions across the main assistants, matched against where your free signups, upgrades and enterprise deals come from, so you fix the gaps that cost the most revenue first. We can build that view with you and plan the fixes: ask us to review what assistants say about your video tool. The work that follows, from consent and rights pages to fresh release pages and fair comparisons, is described on our generative engine optimization service page.

Frequently asked questions

Do AI assistants recommend the AI video tools with the best models?

Not necessarily. No assistant documents how it chooses, and studies of software questions find AI search leans heavily on independent sites. A tool with a strong model but little independent coverage can be well known by name yet rarely suggested, as Sharma’s 99.4% versus 3.32% gap shows.

Should AI video companies publish “alternatives” and comparison pages?

Yes, provided they are fair and state their test conditions. Buyers ask head-to-head and alternatives questions, and vendors in the category already publish such pages. Pages that rank yourself first without test conditions are a weak bet.

Do the EU’s deepfake labeling rules affect AI visibility?

Not directly. They affect what buyers ask. From 2 August 2026, companies publishing deepfake video in the EU must disclose it, so buyers ask which tools support labeling. Clear public answers make it easier for an assistant to answer that question about you.

How can a video company tell whether signups come from AI search?

Use three signals together: AI referral traffic, a “how did you find us?” question at signup, and regular checks of what assistants say for your main video use cases. Referrals by themselves miss part of it, since many AI-driven visits land on the homepage or show up as direct traffic.

Sources

Free strategy call

Some questions are easier to answer about your own business.

Bring the one that matters most. On a free 30-minute call we’ll take a first look at it and send you a short written read afterward.

Ask about my businessHow we work