Set up the prompts
Tags, countries, the questions your buyers actually type. Custom prompts sit next to the ones we seed from search volume. The list is a brief, not a spreadsheet in someone’s Drive.
Solutions·Marketing·AI Search Analytics
What AI answers — and what to change.
AI search analytics for marketing teams. Visibility, position, and sentiment across ChatGPT, Gemini, Google AI Overviews, Perplexity, and Copilot. We show what to improve and what to emphasize. You buy the work from Pattern Automation — the practice sits with marketing.
VisibilityHow often models mention you vs the set you track
PositionWhere you sit when the answer names a shortlist
SentimentHow you are described — not only whether you appear
ActWhat to fix, what to underline, which source to earn
The handoff
The question marketing cannot screenshot from ChatGPT.
Someone asks an AI who to hire, which stack to pick, which firm is serious. The answer cites a review site, a thread, a page you forgot you published. Guessing whether you appear is not a function. Tracking the prompts, the models, the sources, and the next move is. We run this practice on ourselves first. Then we run it for you — as a marketing engagement, not a login you have to babysit.
Pick the winners
Volume tells you which questions are worth owning. You do not track every phrasing. You track the ones that move a deal.
Name the set
Your brand, the competitors, the aliases models already use. The chart is only useful if the shortlist is honest.
Choose the models
ChatGPT, Google AI Overviews, Google AI Mode, Perplexity, Gemini, Microsoft Copilot on every engagement. Claude, Grok, Mistral when the buyer actually asks them.
Find the sources
What the models cite when they mention you — and when they mention someone else. G2, LinkedIn, Reddit, editorial domains. The citation is the lever.
Act on the insight
A profile that needs reviews. A thread to join. A page to underline. Digital PR where the model already reads. We show the move. Then we make it.
Four surfaces, one practice
Overview. Prompts. Sources. Models. Not four tabs you never open.
Typical AI-search tools dump a chart and leave marketing to guess the next sentence. This engagement is a weekly read: what moved, what to emphasize, what to ship.
Overview
Visibility, position, sentiment on one screen. Your line against the set. Filters for brand, window, tags, and model — so a CMO can read it in a minute.
Prompts
The questions, tagged and scoped by country. Seeded from volume, extended with the phrasing your sales team actually hears. Nothing tracked that cannot change a page.
Sources
What the model reads when it answers. We rank the citations, then write the move: a profile, a comment, a page, a pitch. Marketing sees the why. We ship the work.
Models
The assistants buyers already open. One tracker per model is a screenshot. One engagement across the set is how marketing sees the real answer.
Earned · Owned · Impact
Not only whether you are named. How, from where, and whether it helps.
Mention frequency without citations is vanity. We split what you earned, what you already own, and what actually moves the answer.
Earned
Reviews, threads, press the model already trusts. You do not control the page. You can still show up on it.
Owned
Docs, handbook, the site. Pages you can rewrite this week. We tell you which ones the models already cite — and which they ignore.
Impact
The recommendation with a next action: underline this claim, fix this profile, pitch this domain. A chart without a move is a report.
The set
Every engagement tracks the models people actually ask.
ChatGPT mention frequency, citations, sentiment, and a competitive shortlist — plus the Google, Gemini, Perplexity, and Copilot answers that sit next to it.
ChatGPTMentions, citations, sentiment, shortlist
AI OverviewsGoogle’s answer box, cited sources
AI ModeThe longer Google assistant turn
GeminiHow Google’s model names you
PerplexityCited-answer buyers already trust
CopilotMicrosoft’s default in the enterprise
ClaudeWhen the buyer’s team lives in it
GrokWhere the public conversation leaks in
MistralEU-first stacks, same prompt set
Agencies & in-house
Multi-client workspaces. Pitch projects. A weekly read, not a portal login.
In-house marketing gets one brand, one shortlist, a heartbeat. Agencies get rooms per client, a pitch deck from the first snapshot, and reporting that lands in the channel they already use.
Workspaces
One room per brand. Tags, competitors, and models do not leak across clients.
Pitch projects
A first snapshot before the retainer. Enough signal to walk into the room with a chart and a move.
Weekly reporting
MCP into Slack or Teams. What moved, what to underline, what we will ship. Not a dashboard they have to remember.
Branded readouts
The chart in your colours, in the project. Exports when the board wants a file.
How it leaves the room
CSV. Looker. API. MCP. The project stays the system of record.
CSV export
Prompts, visibility, sources — a file the analyst can open. Nothing locked in a vendor screen.
Looker Studio
When marketing already lives in a dashboard, the series lands there. We do not make you learn a second BI.
API
When engineering wants the same numbers in the warehouse. The engagement still has a named owner in marketing.
MCP
A weekly digest into the channel you already read. The agent can draft the note. A person still sends the claim.
How it compares
A ChatGPT screenshot is not a marketing function.
| Typical SEO / GEO tool | Pattern Automation | |
|---|---|---|
| Job | A login and a chart. | A marketing engagement: the chart, the move, and the work. |
| Prompts | You guess what to track. | Volume plus the phrasing sales already hears. Tags and countries. |
| Models | One assistant, or a PDF of screenshots. | ChatGPT, Overviews, AI Mode, Gemini, Perplexity, Copilot — one set. |
| Next step | “Improve your content.” | Earned, owned, impact. Which profile, which thread, which page. |
| Who owns it | SEO, or nobody. | Marketing. Same org chart as campaigns and the yes on the post. |
| Repeat | You remember to open the app. | A heartbeat. Weekly MCP. A person on the claim. |
How it runs
On demand, at the gate, and on the Monday heartbeat.
“Where did we lose the shortlist this week?”
Ask in the project. The session pulls visibility, the sources that flipped, and a ranked list of moves.
Nothing public from a recommendation
A suggested reply, a PR pitch, a page rewrite sits behind Ask. Marketing still owns the sentence.
The week, overnight
A heartbeat re-runs the prompt set, diffs the sources, and parks a digest for Monday. We still do the work the digest names.
Control
A hallucinated mention is still a quote if you repeat it.
Analytics is where an autonomous mistake becomes a claim on a slide. Be precise about the defaults.
Publish is Ask until you say otherwise
Recommended posts, pitches, and page copy wait on a person. The tracker can draft. It does not speak for the brand.
Placeholder names only
Readouts never invent a customer. Harbor Stack and Lumen Grid are illustrations. Your shortlist is yours.
The project is the record
Prompts, diffs, and the yes live in the workspace. A vendor screen is not the system of record.
From the blog
GEO, sources, prompts, shopping — the notes marketing actually uses.
Field notes on AI search analytics. The full set lives on the blog.
GEO vs SEO
Rankings and clicks versus visibility, position, sentiment, and citations inside the answer.
Complete guide to GEO
Grounding, brand mentions versus citations, fan-outs, and a weekly operating rhythm.
Choosing prompts
Volume plus the phrasing sales already hears. Tags, countries, no vanity questions.
Sources
What the model cites, earned versus owned, and the lever to pull next.
AI Overviews
The Google answer box is a property. ChatGPT-only tracking misses it.
KPIs
Visibility, position, sentiment, citations, and pipeline — not sessions alone.
Shopping
Product shortlists in ChatGPT and Gemini: offers, reviews, merchants.
MCP reporting
A weekly digest into the channel marketing already reads. A person still sends the claim.
The same platform
Marketing owns the practice. The other teams still share the chart.
One project, one set of connectors, one memory that compounds. AI search analytics is a marketing role on the same org chart.
MarketingDrafts, then a yes, then the post — this tracker feeds the brief
SalesThe prompts buyers type are the ones the floor already hears
OperationsRoutines run checklists; humans handle exceptions
FinanceCaps, receipts, and reporting packs
SupportAround-the-clock triage with named escalations
EngineeringIssue → PR on heartbeats, QA before merge
HRThe coordination — never the decision about a person
ITRunbooks that execute — and you approve the writes
Data ScienceAn analysis you can re-run — with the query attached
DesignImages, slides, and brand boards as tasks
LegalMatters, citations, and workflows you own
Personal AssistantAI for personal life — Telegram, voice, reminders
“A screenshot of ChatGPT is not analytics. Prompts, models, sources, and a move marketing can ship — then a person on the claim.”
Start with the prompts your buyers already type.
We show what to improve and what to emphasize. You buy the work from us. Contact to scope the first snapshot.