Article · 19 August 2026

How not to get lost in AI slop and choose what works

Marketing activities have tripled, each looks meaningful, revenue hasn’t changed. You should choose not by usefulness — almost everything is useful — but by one question: will this change the number you are responsible for, and over what timeframe.

The problem of modern marketing is not that there is nothing to do. It is that you can do an endless amount, every action looks meaningful, and none of them on its own gives a noticeable result. AI did not create this problem — it multiplied it by ten, because now any activity can be launched in fifteen minutes.

What AI slop is in marketing

Slop is the name for a stream of content produced quickly and in large volume, formally correct and carrying no value. In marketing it doesn’t look like bad copy — the copy is actually decent. It looks like a growing number of activities with unchanged revenue.

A typical picture: three times as many posts, newsletters go out weekly instead of monthly, five landing pages instead of one, twelve new formats invented. Everyone is busy, everyone is happy with the pace. Sales are the same.

This happens because the cost of production has dropped almost to zero, while the cost of audience attention hasn’t changed at all. Previously, volume was limited by resources: writing ten texts was expensive, so you had to choose what exactly to write about. Choice was built into the process by force. Now it isn’t built in anywhere, and it has to be made consciously — and nobody does it.

The three most common traps

The volume trap

If production is cheap, then you should produce a lot. The logic seems flawless, but it holds only for channels where reach depends linearly on the number of posts. Almost no such channels are left: social networks, search engines and AI assistants all rank by engagement and quality, not by volume. In 2026, ten weak pieces deliver less total reach than one strong piece — and on top of that they damage the account’s statistics, reducing impressions for the next ones.

The presence trap

“We need to be everywhere our audience is.” The audience is everywhere — this says nothing about where it makes the purchase decision. Being present in five channels with one person means five channels running at a third of capacity, and none running at full. One channel brought to the point where it generates leads is worth more than five started ones.

The novelty trap

A new format, tool or platform has appeared — we need to try it. The problem is not the experiment, but that experiments are never closed: what has been launched continues to demand time even when it is clear it doesn’t work. A year later the company has twenty activities, fifteen of which nobody has ever evaluated.

A four-question filter

This is what I apply to any list of marketing activities during a review. It takes half an hour and usually strikes out between half and two thirds.

The fourth question cuts the most. It is also the least popular, because it requires admitting that resources are finite and AI has not increased them. AI has sped up production, but it has increased neither audience attention nor the team’s ability to see things through.

How to tell meaning from noise

A practical sign: a meaningful action changes what a person thinks about the product. Noise changes only the number of touchpoints.

A breakdown of a real client objection, published with numbers, changes how people see you. The twentieth post with the advice “calculate your unit economics” changes nothing, even if it is written better than the first nineteen.

Where AI really helps you choose

The irony is that the best use of AI in this situation is not production, but reduction.

• Analyzing inquiries and reviews. Export a year of support correspondence and reviews and ask for them to be grouped by topic with frequency. This gives a list of five to seven questions that really concern the audience — and it always differs from what the company writes about.
• Consolidating actual returns. Put all activities over six months and what came from them into one table. Half the rows will be empty — and this is the most useful document of the quarter.
• Checking for repetition. “Here are twenty of our pieces, find which ones say the same thing.” Duplicates are found at a volume a person can no longer keep in their head.

In all three cases AI works with your data and answers a specific question. This is fundamentally different from “come up with a content plan” — there it gives the market median, which is exactly the noise you are trying to get away from.

Order of actions

1. Write down everything you do. One list, without dividing into important and unimportant. It usually has fifteen to thirty items, and the owner sees half of them in one place for the first time.
2. Run it through the four questions. Each item gets a number, a timeframe and an answer about what you give up.
3. Keep three. Not five and not seven. Three activities seen through to the end give more than fifteen at half strength — and it is the only way to understand whether a channel works or not.
4. Set a review date. For each of the three — a specific date when we look at the number and decide whether to continue. Without a date, an experiment turns into an obligation.
5. Send the rest to the archive, not the backlog. The difference is significant: a backlog creates a sense of debt and comes back at every team meeting. The archive is closed.

What remains once the noise is removed

Usually three things: one acquisition channel that actually brings in leads; one content format that you can do and competitors can’t; and working with those who have already bought. The last one almost always turns out to be the most underrated — by the numbers it is the cheapest source of revenue, and it is exactly what gets lost first when attention shifts to producing new things.

The main thing to understand about AI and marketing: a tool that removes the volume constraint raises the value of knowing how to choose. Previously, the budget chose for you. Now nobody chooses, and this is the only work that remains solely with humans.