Boost an Instagram post through Claude: what the defaults do

A post did well. Put money behind it. In Instagram's own app that is a button, and the app makes you pick a goal on the way through. Through the connector it is one call, and the tool already has an answer for every question the app would have asked.

So the decision is not “boost this post or not.” It is which dozen settings you are content to let a default choose, on a campaign that will sit in your account next to the ones you built by hand.

Three required fields, and a dozen quiet ones

The boost tool declares three required parameters: the ad account ID, the Instagram account ID, and the ID of the post. Its own description says only three fields are required and “all other fields are optional with sensible defaults.” That optional list runs to objective, optimisation goal, destination, targeting, budget, duration, billing event, bid strategy, call to action, special ad categories, and a name for each entity it creates.

The named defaults are specific. Objective OUTCOME_TRAFFIC, destination INSTAGRAM_PROFILE, optimisation goal VISIT_INSTAGRAM_PROFILE. Billing on impressions, lowest cost without a cap, United States targeting, and roughly the five-dollar-a-day equivalent in your account's currency for six days.

Together those make one thing: a profile-visit ad. It sends nobody to a product page. If you are boosting the post because it sells, the default is optimising for something else — and it will do that competently.

The Claude Ads Operator course covers the plan-then-confirm habit this tool is built around.

What you asked for versus what gets built

None of these are bugs. A tool working from three IDs has to fill the rest in from somewhere, and a cheap, short, profile-visit ad is a defensible guess when it cannot know your business. The failure mode is not the defaults. It is assuming the sentence you typed changed them.

The finding-the-post half is not part of the boost

The prompt pattern circulating in vendor guides asks for two things at once. Windsor.ai, a connector vendor writing about its own product, opens an example with “Find my top-performing organic Facebook post from the last 14 days” before the boosting instruction (page last updated 20 August 2026).

On the official connector the second half of that has a tool and the first half does not. The media listing the boost depends on returns advertisable posts, reels and stories with captions, types and permalinks, and its declared filters are a date range, a media type and a product type. No engagement field to rank on, no sort order to ask for. So whatever a model calls your best post, it did not read that here — ask it where the ranking came from. That is the same reflex the setup guide asks for on any number the connector reports.

What is enforced and what is only advised

Two safety properties ship with this tool and they are not equally strong. Everything it makes is made paused, which the description states twice and which no phrasing on your side changes. Budget goes nowhere until you activate the entities.

The plan step is softer. The confirmation flag starts false, so a bare call returns a plan and builds nothing — that much is real. But the instruction never to jump straight to creation sits in the tool's guidance to the model, not in a server-side precondition. Same split as the delete guardrail on custom audiences: one rail is in the server, one is in a paragraph.

A default is a decision someone else made about your money, in advance, without knowing anything about you.

Four things to override before you confirm

1. Destination and optimisation goal START HERE

   If the boost exists to sell, profile visits are the wrong target and the ad set is built around them. Change both together — changing one leaves an ad set optimising for a place it is not sending anyone. The trade: a website destination usually needs a promoted object and a pixel event too, so the one-line boost becomes a three-line one.

2. Targeting

   The United States default is the costliest one for anybody selling elsewhere, and it reads neither your account's country nor your existing ad sets. The trade: the parameter takes a full spec rather than a country name, so this is the field most likely to send you back to Ads Manager to copy one.

3. Budget, in the right unit

   State the amount in minor units and read the plan's currency field back before confirming. A budget you do not recognise in the plan has already told you something is wrong — and the floor it gets clamped to is a number the connector never reports.

4. Special ad categories

   The default is an empty list. If you advertise credit, employment, housing, or social issues, elections and politics, that default is wrong, and the consequence is a policy problem rather than a performance one. The tool infers nothing from your post.

The mistakes

• Passing the shortcode from the post's URL. The tool wants the numeric ID from the media listing and explicitly says not to transform the permalink shortcode. Different strings, same post.
• Reading an empty account list as “no Instagram account.” The listing tool's own guidance says zero results means either no linked account or a missing instagram_basic permission. Two causes, one silence.
• Using an old objective name. Only outcome-style objectives are accepted. The pre-outcome names are not translated for you.
• Confirming a plan you skimmed. It is the only place every defaulted field appears at once. After confirmation you are reading three entities in Ads Manager.

Try this tonight Ask for the plan and nothing else — any recent post, confirmation left off — then read the resolved settings rather than the summary of them. Check the destination, the country, and the currency the budget is in. Anything there you would not have chosen is a field you must name explicitly every time, so write it into the prompt you reuse. Nothing was created, so there is nothing to undo.

This tool is a fast path, not a simple one. It compresses a campaign build into one call by answering a dozen questions for you, and it is honest about which answers it picked. Read the plan, override the four fields above, and it builds what you wanted. Confirm it unread and you have a paused profile-visit campaign aimed at the United States — a perfectly good ad for somebody. Whether that somebody is you is exactly what the plan step exists to tell you. If you are still deciding how much to hand over at all, the list of things to never let it do comes first.