There is a new line item in software marketing budgets: AI search visibility. The category arrived fast, and it arrived in two shapes. One is software you subscribe to that tracks prompts across engines and helps your team produce content. The other is a managed program where an outside team owns the work and reports on the result. Both are sold with the same vocabulary, which makes them look like alternatives when they are actually different purchases.
This guide separates them properly: what each model structurally can and cannot do, the total cost of ownership with the operator counted, the break-even points, and the specific failure modes to plan around. We sell the managed model, so the math is written to be checkable and the section on when to buy software instead is written seriously.
The two models, defined
An AI visibility platform is instrumentation plus workflow. You define a prompt set, the product runs it against ChatGPT, Google AI Overviews, Perplexity, Gemini, and Claude on a schedule, and you get citation and mention data plus opportunity reports. Most also include content drafting workflows, brand knowledge bases, and CMS integrations. Pricing is metered: seats, tasks, tracked prompts.
A managed program is capacity plus accountability. A strategist defines the prompt set, specialists build the pages that answer those prompts, implement the schema and structural changes on the site, do the off-site work that shapes how models describe you, and report through to qualified conversations. Pricing is typically a flat monthly retainer with a minimum term. Ours is published on the pricing page.
Note the asymmetry: the platform sells you the measurement layer and part of the production layer. The program sells you the whole layer, including the surfaces that determine whether the measurement ever improves.
What software structurally cannot do
This is not a criticism of any specific product. These limits are properties of being software.
- It cannot change third-party sources. Generative engines describe your brand from a corpus you do not own: review sites, directories, listicles, comparison pages, forum threads. A tool can show you the sentence is outdated. Fixing it means outreach, placements, and entity cleanup across dozens of properties. That is the bulk of generative engine optimization and it is human work.
- It cannot ship a website. Most AI search recommendations are structural: answer-first page architecture, an entity graph, internal linking, performance, and in many cases a rebuild. A dashboard flags the gap and then waits.
- It cannot decide what matters. Prompt sets encode a positioning hypothesis. Choosing which 20 prompts represent your category is strategy, and getting it wrong produces a clean report on the wrong questions.
- It cannot be accountable. A licence has no obligation to your pipeline. When the number does not move, nothing in the contract engages.
What agencies structurally get wrong
The mirror image is just as real, and worth knowing before you sign anything.
- Vague instrumentation. Many agencies sell AEO in prose with no named engines, no prompt count, and no cadence. If a partner cannot tell you how many prompts they track and how often, they are selling content with a new label. Ours are stated on the AEO service page.
- Metric definitions that drift. Without written definitions of citation share and share of voice, month three reporting quietly changes what it counts. Fix the definitions before launch.
- Slower feedback loops. A tool refreshes on a schedule you control. An agency reports on a cadence, usually weekly written and monthly formal.
- Knowledge that leaves. An internal operator accumulates product understanding that an external partner replicates only partially. That is a genuine long-run advantage for in-house plus tooling.
Total cost of ownership, done honestly
Comparing a licence to a retainer is the most common mistake in this decision. Software only produces output when someone drives it, so the operator belongs in the total.
| Cost line | Platform plus operator | Managed program (mid-tier) |
|---|---|---|
| Licence | $500 to $3,000+ / mo, metered | Included |
| Person operating it | $12,000 to $16,000 / mo loaded | Included, plus specialists |
| Website and landing pages | Scoped separately | Typically included |
| Off-site GEO execution | Not available | Included |
| Paid media management | Not available | Often included |
| Time to first output | 3 to 6 months hiring and ramp | Baseline and first pages inside 30 days |
| Monthly total | $13,000 to $19,000 | $15,000 at our Momentum tier |
The band overlaps, which is the point. At the same spend you are choosing between one person's depth with better tooling, or a team's breadth with less product intimacy. If your website converts and your content engine already runs, the first option is excellent. If either is unbuilt, the second covers ground the first cannot reach at any licence price. The same arithmetic pattern shows up in the broader firm versus in-house comparison.
A decision framework you can run in ten minutes
Score each statement 0 for no, 1 for partly, 2 for yes.
- We employ someone whose job includes organic and AI search, with at least half their week free.
- Our website converts and needs no structural rebuild in the next two quarters.
- We publish at least four substantive pages per month today, on time.
- Our positioning and category language are settled, so a prompt set would be obvious.
- Third-party coverage of us is accurate and reasonably broad.
- We do not need paid media, or it is already well run.
A score of 9 or higher means buy the software: your constraint is visibility, not capacity. A score of 5 to 8 means run a hybrid: keep or buy a licence and buy execution for the two lowest-scoring lines. A score of 4 or lower means a tool will not get used, and a managed program is the honest purchase. If you scored low on every line, the real first project is the website and positioning, not measurement.
The hybrid shape that actually works
Plenty of good programs run both. The pattern that holds up: you own the licence and the data, the partner operates inside your tooling rather than duplicating it, and the retainer excludes monitoring you already pay for. Write three things into the agreement.
- Metric definitions, agreed before the first report, so citation share means one thing all year.
- A prompt set reviewed quarterly, with retired prompts logged, so the baseline stays comparable.
- Data and account ownership in your name, including the prompt archive and any published pages.
Common mistakes in this decision
- Comparing licence to retainer. The operator is the largest line in the software option and it is the one people leave out.
- Buying instrumentation to avoid a positioning decision. A prompt set cannot tell you what category you are in.
- Treating AEO and GEO as one purchase. On-site citation work and off-site representation work need different mechanics, as the GEO versus AEO breakdown lays out.
- Tracking too many prompts. Fifty vague prompts produce noise. Twenty prompts your buyers actually type produce decisions.
- Changing metric definitions mid-engagement. Any trend line becomes meaningless, and it usually happens right when results flatten.
- Buying either one before the site can be edited. If nobody can change page structure this quarter, both options stall.
What to ask each vendor
Ask a platform: which engines, how often refreshed, what happens at the prompt cap, can I export the raw prompt archive, and what does a seat cost in year two. Ask an agency: how many prompts do you track at my tier, on what cadence, what are your written metric definitions, who is the named owner and how many accounts do they carry, and what is explicitly out of scope. The exclusion list is the most informative answer in either conversation, a point we make in the vendor evaluation scorecard.
Where we fit, and where we do not
Momentence is the managed option. We run a fixed prompt set against the five major engines, publish our metric definitions, and own the website, on-site AEO, off-site GEO, content, and paid media as one program at a published flat retainer with no seats, no task credits, and no prompt caps. Our detailed comparison against the platform archetype lives on the AirOps alternative page.
We are the wrong answer if you already have an operator and a site you like, in which case a licence plus that person is cheaper and faster. We are also the wrong answer if your total budget is under roughly $7K per month, where the honest advice is to buy instrumentation and do the work yourself for two quarters.
If you want an outside read on which model your situation calls for, bring your current traffic, your team structure, and your site to a 30-minute growth call. We will score the framework above with you and say plainly when the tool is the better buy.