The best thing about using AI becomes the worst thing about using AI, and the turn happens faster than most teams notice. A tool that removes friction from production starts removing the decisions that production was supposed to carry. The failure is never the model. It is a human preference for being led, combined with a demand for quick results. This is where AI genuinely belongs inside a content strategy, where it must never be allowed, and what changed in my own writing after 14 years of doing it without one.
What does AI actually change in a content strategy?
AI changes the cost of production, not the quality of judgment. It accelerates research, data collection, bulk analysis and drafting. Deciding what a business should say, to whom and what to leave out remains a human responsibility that no model can inherit.
Nothing about that sentence is a warning against AI. Refusing AI is not an option worth debating. This is a commercial world, competitors are using it, and a team that declines the tool on principle is choosing to be slower for no strategic return. The correct move is to include AI and define its work areas precisely. Undefined, it expands into the work it should never have touched.
Why does templatisation turn AI into slop?
Writers using AI drift toward templates very quickly, and the drift is understandable. The model produces a structure that worked once, the structure is reusable, and reuse feels like efficiency. Some weeks later every piece on the site has the same skeleton.
Every topic resists that. Difficulty varies. Intent varies. The queries behind the topic vary, the emotion varies, the persona varies, and the individual person reading varies. A template applied without customisation flattens all six of those variables into one shape, and the result is what the industry now calls slop. The word describes a real failure: content that is fluent, structurally complete and aimed at nobody.
The audit view of this problem is unforgiving, since these pages cluster in the delete band of any inventory review, a pattern I set out in the AI content audit. Templated pages do not fail because a machine wrote them. They fail because nobody decided what each one was for.
Who is at fault when AI writing fails?
The human is at fault. AI is not the problem, and I want that stated plainly, since the industry has spent three years blaming a tool for a management failure.
I have watched brilliant writers get wasted by AI. Not replaced. Wasted, in the sense that their judgment stopped being exercised and then stopped being available. The cause is a human thirst for someone else to lead, plus impatience for quick results. Give a capable person a fluent machine and a deadline, and many will follow the machine rather than direct it.
Consider what the machine is actually doing while being followed. It is predicting the next word from patterns in its training, and where retrieval is involved it is pulling documents into context to ground that prediction. How many people using it daily could explain how retrieval-augmented generation works, or what it does when the retrieved material is thin. Very few. A sophisticated instrument gets operated like a genie, and the wish gets granted in the shape of the average of everything it has seen.
Using AI is not outsourcing your thinking. Take its inputs. AI is not the writer. You are the writer.
What did AI change about my own writing?
The honest numbers from my own desk are less flattering to AI than the marketing suggests, and more interesting.
Before AI, a 1,000-word article took me 10 hours, and at the end of it I could claim with confidence that the piece would rank.
With AI, the same article sometimes takes the same 10 hours, sometimes closer to eight. The time saving is real and modest. What changed materially is my claim percentage: the proportion of pieces I am willing to stand behind as rankable went up.
The sequence is unchanged. I do the research. I decide what goes in and what stays out, paragraph by paragraph, including what each paragraph must say and what it must refuse to say. AI helps write. I read it again, decide again, re-read, and I have lost count of how many passes. It ends when I am satisfied.
Two hours saved on a 10-hour task is not a productivity revolution. A higher hit rate is worth considerably more, since the expensive failure in content is not slow writing. It is publishing something that was never going to work. The gain arrived in the confidence, not the clock, and it arrived only because the decisions stayed with me.
Where does AI genuinely belong?
Define the work areas explicitly, and the tool becomes an asset instead of a slow leak. These are the places it earns its keep.
Building databases at speed, parsing large volumes of source material, gathering what would take days by hand. Data collected by one model can be verified by another, which makes cross-checking cheap enough to actually do.
Clustering at volume, mapping queries, building spreadsheets fast, getting closer to funnel intent across a large query set. Mechanical work at a scale humans do badly and slowly.
Producing wide option sets for a human to select from. The generation is machine work. The selection is not.
Writing to a paragraph-level plan a human has already decided. The model executes the plan. It does not author it.
Grammar checks, consistency passes, UI and UX microcopy and other simple tasks, each one approved before it ships. Useful, bounded, low risk.
Google’s position removes the last excuse for avoiding the tool. Relevant, useful content ranks whether it came from a human, a machine or a collaboration between the two, a point covered in does AI content rank on Google. The method of production was never the question. The value of the output always was.
Where must the human stay?
After templatisation, the second failure is the loss of creativity, and it shows up wherever a piece needed an idea rather than an arrangement.
Bulk research, clustering, spreadsheets, option generation, grammar, interface copy, drafting to a settled plan. Fast, tireless, verifiable by a second model.
The creative call, the editorial insight, the field research, the decision about what to exclude, and the position the business is willing to defend. None of it delegable.
Editorial insight is the irreducible part. Knowing that a paragraph is technically correct and strategically wrong, that a section flatters the company instead of serving the reader, that an argument needs a story rather than another statistic: none of that comes from pattern completion. It comes from having done the work in the field, which is the same reason a strategy built from tools alone fails, as set out in what content strategy actually is. Faster armchair thinking is still armchair thinking.
For a view on where AI belongs in your content operation, or a governed system built around it, the way I work is email-first: send your project details to rajat@rajatjhingan.com and you get a considered reply, not a sales sequence. The engagement model sits on the contact page, and the delivered version on my AI content services page.
Key takeaways
- AI changes the cost of production, never the quality of judgment. Define its work areas or it expands into the work it should not touch.
- Templatisation is the first failure. Difficulty, intent, queries, emotion, persona and reader change with every piece, and a reused skeleton flattens all six.
- The human is at fault, not the tool. The thirst to be led plus pressure for quick results produces blind following of a system predicting the next word.
- A 1,000-word article took me 10 hours before AI and takes eight to 10 now. The real gain is a higher claim percentage, not saved time.
- AI belongs in research, bulk SEO work, idea generation, directed drafting and mechanical checks. Creative calls and editorial insight stay human.
- Google accepts relevant content whether human, machine or collaborative. Production method was never the question.
Rajat Jhingan is a content strategist and copywriter with 14-plus years across SaaS, fintech, edtech, travel and PR. He has built content systems that outranked a million-page competitor on 6,000 keywords and grown a SaaS property past 1.5 million monthly impressions. Email rajat@rajatjhingan.com to discuss a project.
Start a conversation