Two engines. One platform. Pick where you start.
Deliver keeps your clients the answer AI gives, tracked across every engine, fixed, and proven with confidence intervals. Acquire finds the local businesses who are invisible in AI search, so you can win them. Most agencies run both, start with the one you need first.
Scanned live across every answer engine
Switch to annual and save ~17% (2 months free).
Studio
white-labelWhite-label for a small agency book.
billed monthly
- — 125 tracked prompts
- — 5 client seats
- — 5 AI engines tracked
- — 3x3 geo-grid · biweekly refresh
- — Daily AI-Overview trip-wire (10 prompts)
- — 20 on-demand re-grids / mo
- — White-label reports & client portal
Agency
recommendedYour whole client roster, white-labelled.
billed monthly
- — 250 tracked prompts
- — 10 client seats
- — 5 AI engines tracked
- — 3x3 geo-grid · monthly refresh
- — Daily AI-Overview trip-wire (10 prompts)
- — 40 on-demand re-grids / mo
- — White-label reports & client portal
Enterprise
white-labelCustom pool, Grok, daily grid + SLA.
annual, sales-assisted
- — Custom tracked prompts
- — Unlimited brand projects
- — 6 AI engines tracked
- — 5x5 geo-grid · daily refresh
- — Daily AI-Overview trip-wire (10 prompts)
- — Unlimited on-demand re-grids / mo
- — White-label reports & client portal
Prompts are a pool, not credits you burn, track up to the pool, swap any prompt anytime. Agencies get white-label reports + a client portal on Studio and up. Deliver has no free trial; start on Lite and scale as the book grows.
Every number is a real AI answer, with a confidence interval. We never invent a metric.
Each figure on a Deliver report traces to a logged AI answer, the exact prompt, engine, date, and cited URLs. We run every prompt repeatedly and report a 95% confidence interval, and only call a change “confirmed” once the intervals separate. No fabricated “AI search volume,” no black-box visibility score. When we can’t measure something cleanly, we say so, so anyone can check ChatGPT themselves and the report still holds. Honest beats impressive.
- AI presence per engine, the share of local AI answers that name your client, across ChatGPT, Google AI Overview & AI Mode, Perplexity, Gemini, Grok & Claude.
- A 95% confidence interval on every number, multi-run, never a single snapshot sold as ongoing truth.
- Mentioned vs. supported, whether the AI merely names them or actually cites their own site (presence that’s backed, not fragile).
- Share of voice + a Maps Share-of-Local-Voice heatmap, where your client ranks in the Google Maps pack block by block, not just in the AI answers.
- A listings audit, whether your client is present on the directories AI cites (Foursquare, Yelp, Healthgrades, and more), with a submission link for every gap.
- Reviews & reputation, the rating and sentiment AI reads, with a reply drafted for every review that needs one, staged for approval.
- An AI-readability check, whether the AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended) are even allowed to read the site, plus the on-page signals they look for.
- The citation graph, the third-party sources the AIs cite, and the specific ones your client is missing from.
- A ranked fix list, drafted, answer-first pages, a GBP punch-list plus posts and Q&A, review replies, and off-site placements, staged for your approval. Nothing publishes itself.
- Before/after proof, a change is only “confirmed” once the confidence intervals separate, so the lift you hand a client is real, not noise.
- White-label reports + a live client portal (Studio and up), share a link or rebrand the whole thing as your own.
How we measure, honestly
| Answer engines | Up to 7, nativeChatGPT · AI Overview · AI Mode · Perplexity · Gemini · Grok · Claude |
| Localization | Coordinate-pinned geo-griduule + geohash lattice, per metro |
| Maps pack | Share-of-Local-Voice heatmapgoogle_maps, located per pin |
| Listings | Present on the directories AI citeschecked across the AI-cited set |
| Reviews | Rating + sentiment + drafted replieslive Google reviews |
| AI crawlers | Allowed to read the site?robots.txt: GPTBot, ClaudeBot, PerplexityBot… |
| Confidence | 95% Wilson CI, multi-runevery prompt run repeatedly |
| A confirmed change | Only when the intervals separateno single-run blip alarms |
| Every figure | Traces to a logged AI answerprompt · engine · date · cited URLs |
Where a metric can’t be measured cleanly, we say so rather than guess. What never varies: every figure is a real, logged AI answer, engine, date, and cited URLs preserved, and a change is only “confirmed” once the confidence intervals separate.
Exactly what you get, and how.
No credits to budget, no black-box score. Here's the whole measure → diagnose → fix → prove loop, plainly.
How does Deliver actually work, start to finish?+
You add a client brand, its metro, and the prompts a local buyer would actually ask an AI. Each cycle we run those prompts across the real engines, ChatGPT, Google AI Overview & AI Mode, Perplexity, Gemini, Grok and Claude, coordinate-pinned across a geo-grid of the metro, multiple times, and measure how often the AI names your client, with a 95% confidence interval. Pulse then tells you what changed and what it’s worth, drafts the fix (an answer-first page, a GBP punch-list, off-site citations), and stages it for your approval. You approve, it ships, and the next cycle proves the lift. Measure → diagnose → fix → prove, on repeat.
What’s a “tracked prompt,” and is the pool a credit balance?+
A tracked prompt is one buyer question you watch over time, e.g. “best dentist in Austin.” Your plan’s prompt pool is a ceiling you track up to, not credits you burn: track up to the pool, and swap any prompt for another anytime. There’s nothing to run out of mid-cycle and no per-run metering to budget.
How do you measure, and why can I defend the numbers to a client?+
Every number traces to a real, logged AI answer: the exact prompt, engine, date, and cited URLs. We query each engine natively (Google surfaces via the live SERP a consumer sees; the rest via their official APIs with web search on), localize with a coordinate geo-grid so “near me” resolves like a real local buyer, and run each prompt repeatedly to report a 95% Wilson confidence interval instead of a single snapshot. We never fabricate an “AI search volume” or a black-box score. When we don’t have the data, we say so.
What does the fix actually involve, isn’t it just “add schema”?+
No, and that’s the honest part most AEO pitches get wrong. Presence is the symptom; citations are the cause. Pulse shows the exact third-party sources the AIs cite for the businesses they name instead, flags the ones your client is missing, and drafts the work to close them: an answer-first service-area or FAQ page, a Google Business Profile punch-list grounded in the cited competitors’ real categories and review counts, and a ranked off-site placement plan. Schema is hygiene, never the lever we sell. Every draft is staged for your approval, with honest placeholders where a real figure is needed, we never invent a number.
What’s the difference between a brand project and a client seat?+
Brand plans (Lite, Growth, Pro) run Deliver on a single business, your own, or one client, measured in brand projects. Agency plans (Studio, Agency, Enterprise) manage a roster, measured in white-label client seats, with a client-facing portal and reports you can rebrand. Use the Agencies / Brands tabs above to see the ladder built for you.
Is everything white-label?+
On Studio and up, yes, white-label reports plus a live client-facing portal you can share by link or put on your own brand. Enterprise adds a custom prompt pool, the 6th engine (Grok), a daily grid, and an SLA.
Do you guarantee my client will rank in ChatGPT?+
No, and be skeptical of anyone who does. We measure AI-search visibility with preserved evidence and prove movement only when the confidence intervals separate. Outcomes depend on engine, query, geography, and time. What we guarantee is honest measurement, the diagnosis, the drafted fix, and the proof, not a promise no one can keep.
How do I start, and how does billing work?+
Add your first client and prompts, let a couple of cycles run, and Pulse opens with what changed. Plans bill monthly or annually (annual = 2 months free) by your prompt pool + projects/seats + engines + grid cadence. Deliver has no free trial, start on Lite (one brand) or Studio (white-label) and scale as the book grows; the live monitor and proof loop run while you’re subscribed.