Is creative AI really growing the activity of creators? A new Adobe survey conducted with The Harris Poll found that 78% of the US creative AI users surveyed said it helped grow their business or audience faster. The same study revealed a trade-off. Some 68% said they at least sometimes accept output that falls short of what they intended, and 47% of creators who had used AI said they had abandoned the tool mid-project at least once in the past three months because getting the result they wanted took too much effort.
The useful conclusion is therefore neither 'IA guarantees growth' nor 'IA reduces quality'. Rather, these declaratory figures show that a tool can speed up certain steps while moving work to brief, control and retouching. For a creator, coach, consultant, trainer or independent artist, the real advantage is not to produce more at any cost. It consists of reducing technical friction without lowering the minimum level of quality that protects its voice, credibility and offer.
Information verified on 10 November 2026, Meeting Time. The investigation was published on 9 November 2026 and concerns more than 2 000 creators in the United States. It is not a advertised feature or software deployment in some countries, but an opinion study on the uses of creative AI.
What the Adobe study really says
Adobe states that it interviewed, with The Harris Poll, more than 2 000 American creators in October 2026. In this survey, a creator is a person who designs and publishes digital content several times a month to inform, entertain or engage an audience. The sample brings together emerging creators and full-time creators. "Creative AI" covers models and tools that help to imagine, generate or modify images, videos, audio and graphic creations.
78 % applies to respondents who currently use creative AI. They state that these tools have accelerated the growth of their activity or audience. The proportion rises to 89 % among full-time creators. This result is interesting, but the verb counts: respondents declare an acceleration. The study does not publish their income, their statistics before and after, or a comparable group that would not have used AI. It highlights a widely positive perception; It does not demonstrate that AI alone is the cause of growth.
Other responses point to growing pressure to build skills. Sixty percent of creators surveyed believe people with stronger AI skills get opportunities they miss, and 51% expect that gap to widen. Among current creative AI users, 89% say they are confident they can learn the skills they need. Access to a generator is only the starting point. What matters is being able to direct it, recognise weak output and finish work to a professional standard.
Why "go faster" doesn't necessarily mean "go better"
The three most revealing percentages are worth reading together. Some 78% of users credit AI with helping them grow faster. At least sometimes, 68% say they settle for output that falls short of what they had in mind. Among creators who have used AI, 47% say they abandoned it midway through a project in the past three months because reaching the desired result took too much effort. Of that last group, 27% cite the difficulty of changing one part without redoing the rest.
This paradox is familiar: generating a first version can take a few seconds, but get the good version sometimes requires several instructions, comparisons, corrections and manual repeats. Saving time at startup can disappear if the result is difficult to correct precisely. Therefore, the time required is not measured to the first draft. It is measured up to the published file, the script it assumes, the available video or the page that actually allows a person to understand the offer.
This reading extends the VIFLY guide to select an AI tool from a specific deliverable Before comparing promises or demonstrations, we must define the defects that would render the result unusable. A technically impressive but inconsistent image with an artist's identity is not a gain. A script produced quickly but full of trivialities is not ready. A miniature that attracts the click but distorts the promise weakens confidence.
The video reveals the real lack: the ability to deliver
The survey also shows why video remains strategically important. Thirty-eight percent of creators say they produce more video than a year ago, and 65% believe making more video content is important to meeting their goals. Yet 34% of full-time creators say they turned down a paid video opportunity in the past three months because they did not feel able to deliver it.
The most requested aids are very concrete: 21 % would like assistance in adding visual effects or removing unwanted elements, 19 % to mount and organize plans in a complete history, and 18 %] to clean speech or correct annoying sounds. These are not just spectacular generation demands. These are problems of finishing, mounting and control: the steps that separate an idea from a deliverable accepted by a customer.
Example 1: an independent consultant. She can use AI to turn a forty-minute webinar into a first draft for a vertical video, identify three useful excerpts and suggest captions. She still needs to verify the quotations, remove shortcuts that misrepresent her expertise and choose a relevant call to action. If the tool saves her an hour but she publishes a poorly supported claim, the loss of trust outweighs the time saved. Counting generated clips tells her little. The useful measure is how many accurate videos she publishes with a clear next step for viewers.
Example 2: A trainer who receives a sponsored video request. It does not need to become immediately expert on all post-production software. It can entrust to the AI the cleaning of sound or the preparation of a first cut, while maintaining the script, the pedagogical demonstration and the final validation. To accept the mission without overselling its capabilities, it needs a test on an extract, a defined quality standard and a manual or human solution if automation fails. The AIA then reduces a technical barrier; it does not replace delivery competence or responsibility to the customer.
The viewpoint VIFLY : set a quality threshold before looking for speed
The VIFLY perspective is clear: an AI tool is productive only if it increases the number of content completed and assumed, not only the number of variants produced. Speed becomes an advantage when it frees time for decision, proof, relationship and offer. It becomes a trap when it normalizes the "almost good".
It is therefore necessary to define a quality threshold before launching the tool. For a video, this threshold may include intelligible sound, the absence of unverified assertion, a pace adapted to the public and a promise true to the content. For an image, it can impose a credible anatomy, a consistent visual identity, no false text and rights of use included. For an article, it may require primary sources, real examples, identifiable opinion and human rereading. This threshold prevents iteration fatigue from turning an acceptable draft into published content by default.
This discipline complements the VIFLY method for use the AI without producing the same content as everyone. The originality does not come from a magical instruction. It comes from the material provided: experience, observation, proof, taste and editorial choice. Similarly, the article on iIA and the future of creators recalls that a higher volume does not correct a blurred position or an invisible offer. The Adobe study today brings a new measurable warning: even convinced users recognize control compromises and abandonments along the way.
A workable process assigns three roles. A person defines the goal and the criteria. AI speeds up a clearly defined task. A person checks the output and decides what happens next. This fits another finding from the survey. Some 77% of all creators surveyed believe the final creative decision should always be made by a person, including when AI agents are involved.
Nuances and points to watch
First limit: geographical perimeter. The data published on 9 November relate to creators in the United States. They should not be presented as a measure of France, Reunion or Europe. Adobe published in June 2026 another international survey including 2 002 French creators, with close results on some themes, but it is a separate sample, period and questionnaire. Adding or confusing these percentages would produce false accuracy.
Second limit: declaration is not a causal measure. To say that AI "accelerated growth" reflects user evaluation. Without detailed economic data, it is not known what part of the progress comes from the tool, the experience of the creator, its regularity, its market, its budget or other simultaneous changes. The figure therefore does not guarantee any individual result.
Third limit: some calculation bases differ. The 78 % and 68 %] relate to the current creative users of the IA. The 47 % refers to creators who used the AI. The 34 % refers to full-time creators. The agent results only apply to respondents who have already used one for creative work. These populations are not interchangeable.
Fourth limit: Adobe has a commercial interest in the subject. The company develops and sells creative software that integrates AI. The methodology with The Harris Poll and the announced sample size provide a serious framework, but the publication remains a brand communication released on the eve of Adobe MAX. The percentages shed light on the tensions reported by respondents; they do not constitute an independent validation of the performance of Adobe tools.
Fifth vigilance: do not confuse assistance with total delegation. Agents are described as emerging. Among their users, 59 % said that it saves time in total, even by counting instructions, checks, corrections and decisions. This is encouraging, but it also means that supervision is part of the real cost. An agent who works longer without producing a controllable deliverable is not automatically more productive.
What now? A seven-step action plan
- Choose a deliverable, not a general promise. Take a repeated task this week: prepare a short script, clean a sound, decline a cover or structure a newsletter. Avoid judging "IA" on an overly large project.
- Write your quality threshold. Before publishing, list three to five conditions you can verify, such as accuracy, tone, readability, visual consistency, rights, duration or format. Identify a flaw that would require you to revise the work.
- Time until the final result. Count the brief, generations, selection, corrections and audit. Compare this time to the usual process. The first result is not the end of work.
- Limit the role of the tool. Ask for a specific step: clean, structure, propose, subtitle or remove an item. Keep intent, sensitive facts and final validation under human control.
- Set a budget for iterations. After two or three attempts without progress, change your method. Try editing by hand, using a different tool, simplifying the format or bringing in a specialist. This rule helps you avoid getting stuck the way creators did on projects they abandoned.
- Turn blockages into skill plans. If sound, editing or storytelling prevents you from accepting a mission, choose a narrow skill to develop instead of looking for a button that does everything. VIFLY courses for create with a method and prepare useful evidence can help structure this progression.
- Measure a business consequence. Follow what the content triggers: answer, click on an offer, request, reservation or sale. The vIFLY statistics make it possible to distinguish observable visibility, actions and revenues. If the hearing progresses, a LinkHub clear can follow up content logically by combining offers, resources and reservations.
So the right question is not "how many content can I generate?", but "how many useful content can I finish without diluting my point of view?". If the AI shortens one step and makes you more demanding on the next, it becomes a lever. If it pushes you to accept a result you wouldn't have signed otherwise, speed masks a quality debt.
Frequently Asked Questions and Sources
Does creative AI guarantee the growth of an audience?
No. The 78 % correspond to the perception of American users interviewed, not to an experiment proving causality. The AIA can speed up some tasks, but growth also depends on positioning, quality, distribution, trust and supply.
Why do so many creators accept an imperfect result?
The study does not give a single cause. However, it shows that retouching can be difficult: some respondents say that they have to start a lot of work to change a detail. Fatigue, delay or lack of fine control can lead to compromise.
Should we learn the video now?
Video is a useful skill if your audience, offer or clients request it. Start with a precise blocking — sound, editing, narrative or framing — and use the AI as a framed support. The aim is to be able to deliver, not to master all the tools at once.
Can an AI agent make the final decision?
Technically, some systems can sequence several tasks. Editorially and professionally, the final decision must remain attributed to a person who can verify facts, rights, consistency and consequences. This is also the position expressed by 77 % of the respondents.
Do these figures apply to French creators?
No, not directly. The November investigation concerns the United States. A separate Adobe study interviewed French creators in May 2026, but its results should not be merged with those of November.