Get Better SEO Answers: How to Brief an AI

Type a question into Brass-SEO's chat and you're doing something specific: briefing an AI. You probably don't think of it that way. You're just asking about your traffic. But the chat has no meeting notes on your business, no memory of last quarter's conversation, no gut feeling for what counts as good on your site. It works from exactly what you give it in that message and nothing else.

Brass-SEO is built by Copper Sun Content and Creative, LLC. The company also builds Copper Sun AI at coppersun.io, a separate product for marketing teams running full campaigns. Its blog recently published How to Brief an AI Like a Senior Planner, written for marketers handing campaign work to a model. The core idea travels well outside marketing. It applies directly to the plain-English questions you type into Brass-SEO's chat every day.

What Briefing Means (and Why AI Can't Ask Follow-Ups)
The Five-Part Brief Copper Sun Uses
Applying the Framework to Brass-SEO's Chat
Before and After: A Declining-Pages Question
Before and After: A Content Gap Question
When the Brief Still Isn't Enough
Frequently Asked Questions

What Briefing Means (and Why AI Can't Ask Follow-Ups)

Briefing an AI means handing it everything a capable contractor would need before starting the job, because a chat window doesn't pause to ask what you meant. Copper Sun's guide frames this as the same discipline a senior planner uses when handing work to a junior team member, with one difference that changes everything: the junior planner can ask a clarifying question. The AI can't.

That distinction matters more than it sounds. A junior team member who gets a vague brief will come back and ask which quarter you mean, or which product line, or what winning looks like this time. An AI model doesn't do that. It picks the most statistically likely interpretation of your words and runs with it, silently, without flagging the guess.

Brass-SEO's chat works from tools, not instinct. When you ask a question, it can call query_site for your Search Console and Analytics numbers, get_page to fetch a specific page's title tag and content, check_indexing to confirm whether a URL is in Google's index, or get_playbook for one of the advanced tactics guides. Real capability, real data.

But the chat still has to guess at your date range. It also has to guess at which page you mean and what "traffic" is supposed to include. A vague prompt leaves all three guesses to chance, and the model resolves each one toward whatever interpretation is most common, not necessarily the one you needed.

The Five-Part Brief Copper Sun Uses

Copper Sun's guide names five elements a usable AI brief needs, and skipping any one leaves the model filling the gap with a generic default.

Audience in decision context — who's on the other end, described by their current belief and the decision in front of them, not a job title.
A specific objective — the measurable shift or decision the work needs to produce, not a vague goal like "inform."
The brand's positioning claim — what makes this different from the obvious alternative, not generic category language.
Constraints and what to avoid — the terms, angles, and comparisons that are explicitly off-limits.
An output example — one real, approved piece the model can match for format and quality.

The guide's own advice on the first four: "Over-specify; the model ignores what it doesn't need and uses what it does." On the fifth, it goes further and calls a real example the single most useful part of any brief: "One good example is worth more than three paragraphs of style guidance." We touched on this same idea in passing in our content optimization AI prompts post. This is the deeper version.

None of the five elements is complicated by itself. Together they close off the guesses a model would otherwise make without telling you. Name the audience and the model stops writing for a generic reader. Name the objective and it stops padding the answer with context you never asked for. The other three elements work the same way. Each one removes a decision the model would otherwise make on its own. Skip all five and the answer comes back technically responsive to your words and useless for your actual decision.

Applying the Framework to Brass-SEO's Chat

Copper Sun's framework was built for campaign briefs, but four of its five elements map onto an SEO question almost without translation. Audience becomes the page or keyword set you actually mean. Objective becomes the exact decision you're trying to make, not a general mood like "how am I doing." Constraints become your date range and your page scope. The output example becomes the format you want back: a table or a ranked list, with an exact count of results.

The fifth element, brand positioning, doesn't map directly onto SEO analysis the way it does a campaign. What replaces it is precision about the metric. "Traffic" could mean GSC clicks, GA4 sessions, or both. Naming the metric up front does the same job positioning does in Copper Sun's framework: it stops the model from defaulting to whichever interpretation is easiest to compute.

Two worked examples make the gap concrete.

Before and After: A Declining-Pages Question

Brass-SEO's chat can answer a question about falling traffic in seconds, but the answer depends entirely on what you tell it to compare.

Vague prompt:

Why is my traffic down?

This gives the model almost nothing to work with. Down compared to when? Which pages? Down in clicks, in position, or in sessions? The chat will pick a default window, probably the last 30 days against the prior 30, and hand back a broad summary that may not touch the page you're actually worried about.

Well-briefed prompt:

Compare organic clicks and average position for pages under /blog/ over
the last 30 days against the prior 30 days. Show the 5 pages with the
biggest click drop, their current average position, and whether the
title tag or content changed recently. I already checked Search Console
for manual actions — none found.

Every element from Copper Sun's framework shows up here. The scope is narrowed to /blog/ (audience). The comparison windows are explicit (constraints). The output is specified as five ranked pages with two named metrics (output example). Ruling out manual actions in advance also keeps the model from spending its answer restating a troubleshooting step you've already covered.

Brass-SEO's dashboard already runs a version of this comparison automatically through the Brass-SEO Declining Pages button, which checks the last 30 days sitewide. Briefing the chat directly earns its keep when the button's default doesn't fit: a narrower folder, a longer or shorter window, or a page you already suspect.

Before and After: A Content Gap Question

The same gap between vague and well-briefed shows up when you ask what to write next.

Vague prompt:

What content should I write next?

Without a topic, a timeframe, or a threshold for what counts as a gap, the model has to invent all three. It will produce something plausible-sounding and generic, because generic is the safest guess when the brief gives it nothing to anchor to.

Well-briefed prompt:

Using my Search Console data from the last 90 days, find queries where
I have impressions but no page ranking in the top 20. Focus on queries
containing "seo" or "search console." Give me the top 5 by impression
volume, with the exact query and my current average position for each.

The audience here is the reader searching those specific queries. The objective is a ranked list of five gaps, not a mood board of ideas. The constraints are a 90-day window, a topic filter, and a top-20 ranking threshold. The output format is spelled out down to the two data points per row.

The Brass-SEO Content Gaps button runs a similar analysis with its own defaults already set. Typing a briefed version into the chat is worth it when those defaults don't match your situation: a seasonal business that needs a 90-day window instead of 30, or a request scoped to one topic cluster instead of the whole site.

When the Brief Still Isn't Enough

Sometimes a well-briefed prompt still comes back thin, and Copper Sun's guide names three common reasons. The output example doesn't match the format or quality you actually wanted. The positioning stayed too abstract. The audience framing wasn't specific enough. Translated to a Brass-SEO question, that looks like asking for "the worst pages" without naming the metric, or asking for "recent" data without a date range, or asking for content ideas without saying which topic they should cluster around.

The guide's fix applies here without modification: "Fix the input, not the output." If the chat's answer feels off, the faster move is rewriting the question with one more specific detail, not trying to manually edit or extend what came back. A second, sharper question almost always beats a round of back-and-forth patching.

Say the declining-pages example above comes back listing five pages, but two of them are old blog posts you retired months ago and don't care about. That's not a bug in the chat's reasoning. It's a missing constraint. The fix isn't asking the AI to "try again" or apologizing for the mess. It's adding one line: exclude anything published before a certain date, or scope the request to a specific folder that only contains live pages. The second prompt, not a follow-up correction, is where the fix belongs.

Brass-SEO connects to your Google Search Console and Google Analytics 4 data and answers exactly the kind of well-briefed question shown above. Start a three-day free trial and ask it about your own declining pages or content gaps, with the specifics that matter to your site.

Frequently Asked Questions

What's the difference between a prompt and a brief?

A prompt is the sentence you type. A brief is everything that sentence needs to contain for the model to skip guessing: the scope, the metric, the date range, and the format you want back. Short prompts can still be complete briefs if they pack in those details. Long prompts can still be bad briefs if they're vague about all of them.

Do the Brass-SEO quick action buttons already brief the AI for me?

Yes. Each of Brass-SEO's 11 dashboard buttons, including Declining Pages and Content Gaps, sends a pre-built prompt with its own defaults for date range and scope already decided. That's a complete brief for the common case. Typing your own question into the chat matters when your situation doesn't match those defaults, such as a narrower page scope or a different timeframe.

Does a longer prompt always work better?

No. Length isn't the goal, specificity is. A one-line prompt that names the metric, date range, and page scope beats a five-paragraph prompt that never says what "traffic" means. Copper Sun's guide puts it as over-specifying rather than over-explaining: give the model concrete constraints, not more prose.

Can I use this framework with ChatGPT or Claude too?

Yes. Nothing about the five-part brief is specific to Brass-SEO's chat. Audience, objective, constraints, and an output example apply to any AI tool you're briefing, whether you're pulling SEO data or drafting a page from scratch.

What if I don't know the exact date range or page I want?

Say that directly instead of leaving it implicit. "Show me the last 90 days, or a different window if that's not enough data to be reliable" is still a brief. Naming your uncertainty gives the model something to work with, where silence just means it picks for you.