Buy the tool first. In 2026, AI visibility software from Otterly, Peec AI, Scrunch, or Profound costs between $29 and $5,000 a month and tells you exactly which prompts you lose, while an agency costs $2,000 to $15,000 a month and exists to change that answer. Nearly every company that hires an agency before running a tool overpays, because they cannot describe the problem they are buying a solution to. The honest split is simple: tools measure, agencies produce. If nobody on your team is going to write pages, fix schema, clean listings, and earn press mentions, a $189 monthly dashboard just gives you a well documented decline.
The confusion is understandable. Both categories market themselves with the same words, “AI visibility,” “share of voice,” “get cited by ChatGPT,” and both show you dashboards. The difference shows up in what arrives at the end of the month: a report, or published work.
What do AI visibility tools actually do?
They run prompts against AI engines on a schedule and log the answers. That is the core function across the category, and it is genuinely useful. A tool sends your tracked prompts to ChatGPT, Perplexity, Gemini, Google AI Overviews, and Copilot, records whether your brand appears, which competitors appear instead, which sources the engine cited, and how that changes week to week.
The 2026 pricing tiers, publicly listed:
Entry tier: $29 to $200 a month
Otterly starts around $29 a month for roughly 15 prompts across four engines, with a Standard plan near $189. This tier is prompt limited and light on analysis, but it answers the only question that matters at the start: do the engines know I exist?
Mid market: $95 to $500 a month
Peec AI runs roughly EUR 89 a month for 25 prompts and EUR 199 for 100 prompts. Scrunch, acquired by Sitecore in June 2026, starts around $250 a month. This is where competitor benchmarking and source level reporting become usable.
Enterprise: $2,000 to $5,000+ a month
Profound sits here, with custom contracts that third party reviews put in the low five figures annually at minimum. You get large prompt volumes, agent analytics, and integrations. You still get a dashboard.
Free and native options belong in the stack too, and most buyers skip them. Bing Webmaster Tools publishes an AI Performance report showing Copilot citations at no cost. Google Search Console surfaces AI Overview impressions. Google Analytics 4 added an AI Assistant default channel group in May 2026, so referral traffic from chatgpt.com, perplexity.ai, and gemini.google.com is now visible without custom regex. Between those three, a competent marketer can build a baseline for $0.
Not sure what the engines currently say about your company? Get your free AI visibility audit and see your live citation picture across ChatGPT, Perplexity, and Google AI before you spend anything on software or services.
What can a tool never do?
Produce the assets that change the answer. Every AI visibility platform on the market is read only against the engines, and citation change requires write operations against your site, your listings, your schema, and third party publications. Specifically, no tool will:
1. Write the content engines cite
Citations go to pages that answer a question directly, with numbers, in structured form. Somebody has to write those. Tools surface the gap, then stop.
2. Fix the technical layer
Blocked crawlers in robots.txt, JavaScript rendered content GPTBot cannot read, missing FAQPage or Organization schema, slow or broken pages. These are engineering tasks.
3. Clean up entity data
NAP inconsistencies across Google Business Profile, Yelp, Bing Places, Apple Maps, and industry directories. This is manual reconciliation work, sometimes across dozens of records.
4. Earn third party mentions
Engines corroborate. Reddit leads citation share across ChatGPT, Gemini, Perplexity, and AI Overviews per Peec AI’s 30 million source analysis, YouTube carries roughly 23.3 percent of Google AI Overview citations per 5WPR research, and LinkedIn climbed to a top five ChatGPT source by February 2026 in Profound’s tracking. Getting mentioned in those places is outreach and PR labor.
5. Decide what to prioritize
A dashboard reporting 40 losing prompts does not tell you which three to fix first. That is judgment.
When is software alone the right call?
When you have execution capacity and no diagnosis. Three profiles where buying a tool and skipping the agency is correct:
You have an in house content team already publishing weekly. Add a $189 to $500 tool, point the existing team at the gaps, and you have functionally built an internal AEO program for the cost of software plus reallocated hours.
You are early and testing. A pre revenue or small company should spend $29 to $95 for three months, learn whether AI search drives anything in its category, and decide with data rather than vibes.
You are large enough to have a specialist. If you can justify a full time AI search hire, the loaded cost typically runs $85,000 to $140,000 a year, roughly $7,000 to $11,700 a month, plus $500 to $2,000 in tooling. That combination beats most agency retainers on volume, assuming you can hire someone who has actually done the work, which in a discipline this young is the hard part.
When does an agency earn its retainer?
When the bottleneck is production, not information. Agency retainers in this category run roughly $2,000 to $15,000 a month depending on scope, and the honest test of whether one is worth it is what ships. A retainer that produces a monthly report plus “recommendations” is a tool with a markup of several thousand dollars. A retainer that produces published pages, deployed schema, reconciled listings across every directory, and earned mentions in publications engines actually retrieve is buying labor you would otherwise hire for.
Ask for the deliverable list in writing before signing. The scope that justifies a real retainer usually includes a technical AEO audit and fixes, a defined number of published or rewritten pages per month, schema implementation, entity and listings cleanup, digital PR or placement work, and monthly citation reporting, which is where the agency’s own tool license lives. If the proposal cannot name a count for pages and placements, the retainer is a subscription to meetings. Buyers evaluating scope should compare proposals against what a full service AI visibility engagement actually includes rather than against each other’s slide decks, because agency decks are remarkably similar and agency output is not.
The most common failure mode is buying an agency for measurement. Firms do this constantly: they pay $5,000 a month, and the agency’s core deliverable is a rebranded export of a $250 dashboard. Measurement is the cheapest part of this category. Pay for production.
What does the honest cost comparison look like?
Three paths, priced over twelve months at realistic mid market scope:
Tool only: $29 to $500 a month in software, $350 to $6,000 a year, plus whatever internal hours you redirect. Zero if nobody executes.
Tool plus in house specialist: roughly $90,000 to $150,000 a year fully loaded. Highest ceiling if you hire well, highest risk if you hire wrong in a discipline barely three years old.
Agency retainer: $24,000 to $180,000 a year. Buys a team instead of a person, buys existing publisher relationships you cannot build quickly, and shifts execution risk off your headcount. Worth it exactly to the extent the contract specifies output.
Timelines are the same in all three cases, which is worth internalizing before anyone promises otherwise: Perplexity retrieves live and can reflect new content in one to two weeks, ChatGPT’s Bing backed index typically takes 2 to 6 weeks, and Google AI Overviews follow organic rankings. Nobody, software or agency, moves those windows.
FAQ
Should I buy an AI visibility tool or hire an agency first?
Buy the tool first, in almost every case. A $29 to $250 monthly subscription from Otterly, Peec AI, or Scrunch establishes a baseline in two weeks and tells you which prompts you lose and to whom. Walking into agency conversations with that data changes the negotiation entirely, because you can demand a proposal against specific gaps instead of accepting a generic scope.
What is the cheapest way to start measuring AI visibility?
Free. Bing Webmaster Tools publishes an AI Performance report covering Copilot citations, Google Search Console shows AI Overview impressions, and GA4’s AI Assistant channel group, added in May 2026, tracks referrals from ChatGPT, Perplexity, and Gemini. Manually running twenty of your buyer’s real questions through each engine once a month costs an hour and catches most of what a paid dashboard would.
Do agencies just resell the same tools?
Many do, and that is fine as long as it is not the deliverable. Most agencies license Profound, Peec AI, or Otterly and include reporting in the retainer, which is reasonable because you would otherwise pay for it separately. The problem is retainers where reporting is the whole product. If the monthly output is a dashboard export and a call, you are paying agency rates for a $250 subscription.
Can one tool track all AI engines?
No tool covers everything, and coverage varies more than pricing pages suggest. Most platforms track ChatGPT, Perplexity, Gemini, and Google AI Overviews; fewer cover Copilot, Claude, Grok, or DuckDuckGo AI. Research across 680 million citations found only 11 percent of domains are cited by both ChatGPT and Perplexity, so single engine tracking will badly misrepresent your position. Check the engine list against where your buyers actually are before subscribing.
How much should a small business spend on AEO in total?
Realistically, $500 to $3,000 a month all in for a local or small service business: a $29 to $189 tool, plus either internal hours or a small retainer covering content and listings work. Below $500 a month with no internal execution, expect measurement and nothing else. Above $5,000 a month, insist on named deliverable counts, because at that level you are buying production capacity and should be able to count it.
What is the single biggest waste of money in this category?
Paying an agency to tell you what a tool already told you. The second biggest is buying enterprise software at $2,000 to $5,000 a month while publishing nothing, which produces a precise, well visualized record of losing. Both mistakes come from the same confusion: measurement feels like progress, and it is not.
Software and services solve different problems, and the sequencing matters more than the vendor. Measure first, cheaply, until you can name the specific prompts you lose and the sources winning them. Then decide honestly whether anyone on your team will do the writing, fixing, and outreach that changes those answers, and buy the missing half. Companies that skip the diagnosis buy the wrong half and conclude the whole category does not work. Start with the diagnosis and skip the guessing: request your free AI visibility audit and get a prompt by prompt picture of where you stand before a single contract gets signed.
Tagged