Track AI Visibility In-House. Should You Pay Someone to Fix It?

The question people are actually arguing about

Your competitor appears in an AI answer. You do not. Do you build a tracker and fix that gap in-house, or pay someone to handle it? Smart operators disagree because the two jobs hide very different costs. Monitoring can be cheap enough to run from a spreadsheet. Getting your brand into more of the answers that matter takes content, technical work, and attention that a busy team may not have.

Most businesses buy the wrong half. They pay for a dashboard, then keep the execution work they never get to. Or they avoid the retainer, congratulate themselves on a low-cost tracker, and watch its list of missing citations grow. Measuring visibility and improving it are not the same purchase.

I spent years managing search budgets where clients bought reporting software they barely opened while the campaign work that paid the bills went short. So I want to give both sides their best case. First: why a small business should keep AI visibility work in-house. Then: why it should pay an external partner. After the costs and the rebuttal, I will rule on the one consideration that decides it for me.

The case for doing it in-house

Your customers can choose better prompts than a generic dashboard

The strongest DIY argument starts with the questions, not the software. You do not need an enterprise license to ask an AI engine which businesses it recommends. An in-house team can assemble 20 to 50 commercial problem-and-solution prompts, such as ‘best inventory sync software for Shopify’ or ‘why our dental billing keeps getting denied,’ then record the answers and citations in a Google Sheet. Practitioners testing multi-model visibility report that running a fixed prompt set three times per query helps smooth out variation for pennies.

The prompt list is where the in-house team has an advantage. Its people hear buyer objections and product questions directly. As our guide to measuring AI search visibility without trusting one screenshot explains, specific commercial pain-point prompts generate 2.4 times more brand mentions than generic category keywords. A tracker built around real customer language is more useful than a polished report on questions nobody asks.

Own the prompt set. The person choosing the questions has a hand in deciding which gaps the business notices.

A schedule, a sheet, and API credits can cover the monitoring

The mechanics do not justify a large retainer on their own. With n8n or Make, an API key, and a weekly schedule, an operator can send prompts, check whether the brand or domain appears in the answer or citations, and timestamp the result. DataForSEO’s SERP API extracts Google AI Overviews and cited URLs starting at $0.0006 per query, or $0.60 per 1,000 queries. One community-built workflow puts a weekly tracker at roughly $6 a month in API credits.

That is not the only price on the market. Semrush lists its AI Visibility Toolkit at $99 a month per domain as an add-on, with daily tracking capped at 25 custom prompts. An Ahrefs Brand Radar review describes add-ons that push the software bill to $828–$1,148 a month per domain. A homemade sheet still incurs API usage and maintenance, but it does not make you buy another subscription tier to try a different prompt.

Do not buy a retainer for a screenshot

Paying an agency purely to monitor AI mentions can mean paying four figures for a slide deck of work you could automate. In traditional search, rank tracking called for proxy pools and complicated results-page parsing. For a defined set of AI prompts, the basic monitoring task is to run queries and record what comes back. A $3,000 monthly ‘monitoring’ package may leave you with little more than an automated workflow’s output pasted into slides.

The in-house advocate’s conclusion is hard to dismiss: keep control of the questions, keep the monitoring budget small, and spend the rest on work that might change the answers. If your team can do that work too, an agency has a much weaker case.

The case for paying someone

A tracker cannot clear Tuesday’s backlog

The strongest case for outside help begins the morning after the report arrives. Your script logs 35 queries where a competitor appears and you do not. What happens at 9:30 on Tuesday? Who investigates the sources, rewrites the comparison page, or takes on the technical fix? The marketing coordinator has a launch. The developer has a sprint backlog. The founder has payroll.

An AEO agency teardown calls monitoring without action paying for a thermometer that only reminds you of the fever. That is blunt, but fair. A visibility score is not progress until someone can act on it. A spreadsheet showing an 8% share of voice in ChatGPT is otherwise a recurring reminder of unfinished work.

The work extends beyond your own site

Improving AI visibility is not a matter of changing a few prompts or adding markup and waiting for citations. In one analysis of 1,885 pages that added schema markup, Ahrefs observed no meaningful citation uplift. Approximately two-thirds of ChatGPT’s cited sources live on third-party aggregators, industry directories, and Reddit discussions. You cannot edit those from your CMS.

That leaves a substantial queue: audit what outside sources say about you, publish thorough answers to buyer questions, refresh comparison pages, and build authority beyond your own domain. Technical structure matters, but it does not replace the content and distribution work. An in-house team handling all of that alongside its usual responsibilities may produce three articles and then stop.

The outside partner’s real value is capacity to ship, not a secret prompt or a prettier chart.

Someone has to stay on the job in month four

Month one of an internal tracker feels productive: the automation runs, the sheet fills, and the team finds gaps. In month two, it writes two posts. In month three, a product issue takes priority. By month four, an API change has broken part of the sheet, the error notifications are unread, and nobody has returned to the missing citations.

Outside help is not immune to that failure. Agency veterans acknowledge the month-four drop-off when a retainer delivers diagnostic frameworks rather than continuing execution. But that is the best argument for paying a capable partner: you are buying an operating rhythm that survives the next urgent internal project. Judge the partner by what it publishes, fixes, and follows through on, not by how often it presents findings.

What each path really costs

The listed cash range runs from $72 a year for a $6-a-month DIY API loop to $65,000 in year one for a $5,000-a-month agency retainer plus a $5,000 audit. That range does not make DIY free or an agency effective. It describes invoices, not the value of work completed. Here are the recurring charges, one-off charges, and costs that tend to stay out of the proposal.

The DIY row is cheap in cash because it prices monitoring, not the work required to improve visibility. Its staff-time estimate is an opportunity cost, not another vendor invoice. If your team already has the time and skills, that trade can make sense. If it does not, the small API bill is beside the point.

Subscription pricing can obscure a different limit: how many questions you can ask. Otterly.AI starts at $29 a month for 15 search prompts across four base engines and rises to $189 a month for 100 prompts; Gemini and Google AI Mode are separate monthly add-ons. Profound AI’s pricing review describes earlier public tiers of $99–$399 a month before a move toward custom contracts. Prompt caps, engines, and add-ons drive the bill. None of those charges writes a page for you.

Agency pricing can obscure what happens after the audit. WebFX lists small-business retainers at $1,500–$5,000 a month, or $18,000–$60,000 a year before a possible initial audit. That fee might buy sustained delivery, or it might buy recurring diagnosis while production remains on your team’s desk. Ask what gets shipped and who owns it.

The cheapest mistake is a $200-a-month dashboard nobody opens. The most expensive is a $4,000-a-month agency that hands you the same dashboard in a monthly PDF.

Where both sides get the job wrong

The DIY camp can mistake discovery for progress. Finding that an answer omits your brand takes seconds. Earning a useful mention may require a better answer to a buyer objection, a stronger comparison page, clearer technical data, or work on sources you do not control. A low-cost tracker tells you where to look. It does not make room in anyone’s week to do the work.

The agency camp can make the opposite mistake at a higher price: package that discovery as an ongoing strategy retainer. I have seen the equivalent in PPC: a percentage-of-spend fee attached to mechanical account checks while the important campaign work waits. A monthly deck of AI answers is the same old trick in a new font. Pay for delivery you cannot staff, not observation you can automate.

My verdict: measure in-house, then buy the work you cannot ship

The deciding consideration is execution capacity. Keep measurement in-house: choose a fixed set of commercial prompts, run it on a schedule, and inspect the answers and citations. Start with a 30-prompt panel in a Google Sheet or a raw SERP API. You do not need a retainer to establish where your brand is absent.

Then ask a less comfortable question: can your team publish two deeply researched, citable pieces a week while managing outside citations and technical structure? If it can, keep the work in-house. If it cannot, buy execution rather than another way to display the backlog. For a business with meaningful search spend and an already crowded marketing queue, that means an execution engine rather than a diagnostic agency. groas takes that route: purpose-built models execute paid and organic search work continuously, with a named human strategist setting direction and answering for attributable revenue.

An early-stage business with under $10,000 a month in search revenue at stake is a different reader. If its founder or marketing lead likes writing technical documentation and its developers have spare sprint capacity, a $6 monthly script and self-written comparison pages can preserve cash and control. I would keep that work in-house.

But if your team cannot reliably update Google Ads negative keywords twice a week, do not build a plan around auditing twenty community threads and producing five technical comparison frameworks this month. Keep the cheap measurement. Put guardrails around the work you buy. Then make sure something changes in the answers, not just in the dashboard.