How Many Blog Posts a Month Do You Need for AI Search Visibility?

Most early-growth B2B companies need 8 to 16 planned posts a month to build real AI search visibility, plus an on-demand reserve of roughly a quarter of that number sitting on top of the baseline. Twelve planned and five reactive is a sane starting point. Thirty and ten is the pace for a company trying to own a category inside two quarters. Below eight planned posts a month, coverage accumulates slower than your buyers change the questions they ask. Above 25 a month without a real editorial system, quality slides and thin posts start working against you.

The number is the easy part. The split is what most content plans get wrong.

What does "planned" versus "on-demand" actually mean?

A planned post is a slot decided before the month starts. Topic, format, and publish date are set in advance. Planned posts build coverage: a body of work across the questions your buyers actually ask, structured so an engine can lift a paragraph out of it and attribute the answer to you.

An on-demand post is one you decide to run this week, for a reason that did not exist last week. A competitor shows up in an answer where you used to appear. A prospect asks the same thing in three sales calls running. You ship a feature. You read a report and have a real opinion about it. On-demand posts are how you answer the market in days instead of next quarter.

Both are blog posts. They do completely different jobs, which is exactly why they should not draw from the same budget.

Why should the two be separate pools instead of one number?

Because when they share one pool, the reactive posts eat the baseline. Anyone who has run a single monthly content number knows the pattern: three urgent posts land in week two, and the planned calendar quietly slips to next month. Do that twice and you have no coverage, just a pile of one-offs.

Clarity's blog agent treats volume as two independent monthly pools for that reason. Scheduled posts and on-demand "Run now" posts draw on separate allowances, so a reactive post can never consume a scheduled one. Each pool shows a live meter with used-of-cap, a progress bar, and its reset date. When a pool is spent, the behavior is graceful instead of a hard failure: the meter turns amber, the copy reads that scheduled posts are paused until the reset date and that on-demand runs are unaffected, and an inline upgrade action appears. It works the same way in reverse. An on-demand run over its cap gets refused with a message saying the scheduled posts are untouched. Both pools reset at the start of the calendar month.

That separation is the difference between a reactive month and an empty month.

How much volume do you actually need, and for how long?

Volume alone does not earn citations. Coverage does. An answer engine assembles a response from whichever sources address the question most directly, so what you are building is a body of work that answers every real question inside one subject area, not a stack of loosely related articles.

Our working calibration for a narrow topic area: 30 to 50 substantial posts, each one fully answering a single question. At 12 planned posts a month you are there in about a quarter. At 4 a month you get there in a year, and by then the questions have moved.

Two things follow.

• Consistency beats bursts. Twelve posts a month for six months compounds, because each new post lands on a site that already covers the neighboring questions. Seventy posts in one heroic sprint, then five months of silence, does not.
• Repetition is the real risk at volume. The failure mode at 10 or more posts a month is not running out of ideas. It is writing the same three ideas in slightly different words. Fix it before the month starts: write the exact question next to every slot, and the format it will take. If you answered a question in the last 30 days, a reworded version of it is not a new post. Nobody cites the second copy.

What cadence should you publish on?

Cadence is the rhythm your planned pool spreads across, and it should follow the pool size, not your ambition.

• Weekly or monthly fits a pool of about 12 planned posts, up to 2 posts per week. This is where most companies under a few million in ARR should start.
• Daily needs at least 28 scheduled posts a month to fill the schedule. Below that you cannot honestly run it, and it is the wrong goal anyway until broad category coverage is the actual objective, which usually means a funded content program or an agency running several sites.

Whichever you pick, hold the reserve. Five on-demand runs a month covers almost anything the market throws at you without touching the baseline.

How do you know the volume is working?

Everyone counts posts published. Count the posts that produced a lead.

That is not a slogan. It is the only way to size next month's pool honestly. If eleven of your twelve posts last month generated nothing and one brought in four qualified leads from Perplexity, the answer is not "publish 20," it is "publish 12 more like that one." Volume without attribution just makes the guess more expensive.

The economics are worth the discipline. Semrush found LLM-referred traffic converts 4.4x higher than organic, G2 reports 71 percent of B2B buyers now use AI chatbots for software research, and Search Engine Land clocked 527 percent year-over-year growth in AI search referrals. Fewer visitors, better ones. So a post an assistant cites is worth measuring on its own, not averaging into a monthly traffic line.

This is the part we built Clarity around. We tie an AI-engine visit to the post it landed on and the lead that followed, with revenue down to the dollar where your store or CRM passes a value, and a projection where the business is leads-only. Bosten Shoes went from unmentioned to the number-one recommended leather shoe brand in El Salvador and logged its first ChatGPT-attributed sale in under 30 days. That is the useful shape of a result: not a rank, a sale you can trace back to a page.

Where should the planned topics come from?

Once you know what to write about, the volume question is trivial. Getting the topics wrong is what makes 12 posts a month expensive: 12 chances to write into the dark.

The honest source is evidence. Which buyer questions do the AI engines answer with a competitor's name instead of yours, and which of your existing pages already earn AI traffic worth deepening. Those two lists fill a planned calendar for a quarter without a single brainstorm.

If you want to see where you stand before committing to a cadence, the free AI visibility check is one snapshot of which questions return competitor names instead of yours. For how the planned pool gets filled week to week, we wrote up building an AEO content calendar that closes visibility gaps, and the pricing page lists the exact scheduled and on-demand allowances per plan.

Pick a baseline you can hold. Protect it with a separate reserve. Then judge it on leads, not on the count.