Free content idea generator for a stronger first draft.

Take Northgate Books, a two-person bookkeeping practice in a market town that works almost entirely with independent cafes, delis and takeaways. The inputs are a subject of hospitality bookkeeping, an audience of owner-operators running one to three sites, and a next step of a twenty minute call about VAT on food and drink. Nothing else goes in: no keyword list, no competitor blog to react to.

What comes back is a set of angles sorted by where each one came from, rather than a list of titles. From questions clients ask: what counts as zero rated when a cafe sells the same slice of cake hot and cold; why the first quarter after a till change is the one where the numbers stop matching; what to do with a shoebox of card machine slips three weeks before year end. From objections: whether a practice this small can cope with three sites; whether moving off a spreadsheet mid year creates more work than it saves. From proof: a walkthrough of one month's reconciliation for a two-site takeaway, with the four places a mismatch usually hides. From process: how a fifteen minute weekly routine replaces the January scramble. From opinion: why most hospitality owners watch the wrong margin, and which one to watch instead.

These hold up because each angle names a decision a reader is standing in front of. The hot and cold cake question is a real VAT boundary that a cafe owner has argued about with somebody. The till change quarter is a specific moment rather than a general worry. Each angle also arrives with its evidence attached: whoever writes it already has the reconciliation, the shoebox and the argument, so the draft comes out of memory in an afternoon instead of being assembled from other people's articles.

The sorting matters as much as the angles. Grouped by question, objection, proof, process and opinion, the shape of the month is visible at a glance. Five questions and no opinion gives you a month that answers well and is forgotten. Four opinions and no proof gives you a month that argues without ever showing the work.

The weaker version is what a blank calendar produces: bookkeeping tips for small businesses, why cash flow matters, the benefits of cloud accounting, common bookkeeping mistakes. Those are categories wearing a topic's hat. None of them names a reader, a moment or a claim, so the draft has nowhere to start and ends up as a definition followed by five bullet points that could carry any practice's logo. They also compete with every accountancy firm in the country publishing the same four titles, while the hot and cold cake question competes with almost nobody and is the piece a cafe owner would forward to whoever runs their kitchen.

Skip this while the last three ideas are still unpublished. A backlog of unwritten topics is not an idea shortage, and adding to it makes the next planning session slower rather than easier. If drafts are stalling, the constraint is time, confidence or a missing reviewer, and a longer list treats none of those.

Skip it when the pieces you already published were not read. Ideas are one input to that, and usually not the broken one. Before generating more topics, check whether anyone knew the last piece existed: where it was posted, who it was sent to, whether anything you control links to it. A better idea distributed the same way tends to reach the same nobody.

Be careful using it on a subject you have no first-hand contact with. The output can only recombine what you feed it, so if you have not spoken to a customer in that field you will get plausible topics you cannot write with any authority, and the draft ends up stitched together from search results. Ten conversations are the useful work there, not thirty angles. The same caution applies to a piece that has to rank: look at what already exists for the query before committing a slot, because an angle that is genuinely good and genuinely covered by three thorough pages is still a poor use of the week.

And skip it for one-off pieces whose job is fixed by an event: the launch announcement, the price change note, the recruitment post. The content of those is already decided. What you need is one clear draft, and choosing between five angles for an announcement is mostly a way of postponing it.

An idea recorded as a noun phrase decays fast. Pricing, onboarding, seasonality: three weeks later you cannot recover what you meant, and the slot goes to whatever is easiest that morning. A usable record has three parts: who it is for, the specific claim or question, and what the reader should be able to do differently afterwards. Compare pricing with how to tell a customer on a rolling contract that the price is going up, for owner-operators, so they send one email instead of losing the account. The second one can be drafted by somebody else on your team.

Keep the source attached as well. An idea that came out of a support ticket is worth more than one that came off a competitor's blog, and in a month you will not remember which was which unless you noted it. Attach the evidence you already hold: the screenshot, the invoice, the two sentences the client actually said. Ideas die in the gap between capture and drafting, and most of that gap is the cost of rebuilding context you had at the time and did not write down.

A keyword is demand. It tells you people type something, not what you have to say about it. A headline is a promise made once the piece exists. The idea sits between them and is the only one of the three that decides whether the draft is worth writing, because it is a claim plus an audience plus a reason the claim is not obvious. You can hold a strong keyword and no idea, which produces the overview page nobody finishes. You can hold a strong idea and no keyword, which is fine for a newsletter and a poor bet for a page you want found.

In practice this means resisting the urge to turn the list into titles early. A title fixes the framing before you know what the piece argues, and once a title exists people write to fill it. Hold each item as a sentence describing what the reader will believe afterwards. Write the headline last, when you know which part of the draft turned out to be the interesting one, which is often not the part you expected when the idea went on the list.

One idea should produce several pieces, though not by pasting the same text into each place. A newsletter can start in the middle, because the reader chose to be there and knows who you are, so open on the specific incident and let the general point arrive late. A search led page has to establish the question in the first two lines, because the reader came from a list of ten options and is deciding whether to go back. In a social feed the idea has to survive with no surrounding context, which usually means taking the single most arguable sentence and letting the rest go. A short video needs the claim in the first few seconds and no preamble at all.

The test is which part of the idea you would lose first. If an angle only works at 1,200 words with three worked examples, it is a page, and forcing it into a post produces a stub that satisfies nobody. If it collapses into one sentence with nothing underneath, it was a social post, and stretching it to a page adds padding a reader can feel by the third paragraph.

A generated list is raw material and should be cut hard. Take each angle and ask two things: can you write this from what you already know, and would a specific person you can name want to read it. Anything failing both goes. Anything failing only the first becomes a research task with a date on it, not next week's post.

Then sequence deliberately instead of working down the list. Alternate the sources so a month holds a question piece, a proof piece and something you are prepared to argue about. Angles produced in one sitting tend to sit at a single level of abstraction, so publishing them in the order they arrived gives you a run of pieces that read like chapters of the same essay. Put the one you are least sure about early, while there is still time to learn from how it lands.

Keep the offcuts. An angle rejected in August is often exactly right in November, when a client asks the question that makes it live. A rejected list is a cheaper starting point next time than an empty page, as long as you kept the reason each one was rejected.