01 · Measure
See exactly how you show up
Every assistant, every query in your campaign. Where you appear in the answer, how you're described, and who gets cited instead of you.
The Strategy Engine
Dashboards tell you where you stand. The Strategy Engine changes it.
HeyOtis finds the move that grows your AI recommendation share, verifies it actually shipped, and proves whether it moved the metric - a campaign-led loop that compounds, not another dashboard.
Why now
AI search created a new channel most teams can't manage
Dashboards without direction don't drive results.
For years, discovery meant ranking on a page of links. AI assistants changed the shape of the journey - buyers now ask what to choose and act on a single recommendation.
Your brand can be strong in search, strong in retail and well known in market, yet still be missing, misrepresented or outranked when an assistant explains the category. Measuring that is table stakes. Closing the gap - and proving you did - is the hard part.
The inputs
Five signal streams. One picture of how AI sees you.
The engine is only as good as its evidence - so it ingests the evidence. Your answers, your traffic and bot logs, your analytics, your own pages and your competitors' wins all flow into one model of the gap.
AI answer sampling. How five assistants answer, cite and rank you across the prompts that matter.
AI traffic & bot logs. GPTBot, ClaudeBot, PerplexityBot, ChatGPT-User - which pages they fetch, and the humans assistants send you.
Site analytics. Sessions, conversions and landing pages, so lift ties to business outcomes.
Your surfaces. Crawls of your own site - structured data, freshness, what actually shipped.
Competitive signals. Who wins the answer when you don't, and why.
Every signal, one model
Answers, bot logs, analytics and your own pages - together
The engine is only as good as its evidence, so all five signal streams flow into one model of the gap.
The loop
One campaign-led loop, end to end
Most tools stop at Measure. The Strategy Engine closes the loop - verifying the move shipped and proving it changed how AI recommends you.
02 · Diagnose
Find the reason, with evidence
Detectors read your surfaces and the answers, and every finding traces to a fact you can click: the answer, the citation, the page, the date.
03 · Prioritise
Get the short list that matters
Ranked by impact and effort, scoped to a campaign, sized to get done. If we can't verify it, we don't recommend it.
04 · Verifywhat others skip
We watch the work land
The engine re-checks your site until the change is live. Nobody has to tell us. If it slips later, you'll know that too.
05 · Provewhat others skip
Measure the lift
Visibility, referrals and conversions, tied back to the move that earned them.
How deep it goes
Four levels the engine operates at
From telling you what's happening to doing the work for you.
01
Diagnostic
Here's what's happening
Where you stand across every assistant, and the signals explaining why.
02
Prescriptive
Here's what to do about it
A ranked, evidence-backed action plan - the moves with the best return first.
03
Predictive
Here's what's about to happen
Emerging prompts, drift and competitive risk surfaced before they cost you the answer.
04
Autonomous
We've already done it for you
The engine ships and verifies the fix, then proves it moved your recommendation share.
The action plan
From signals to the moves that matter
Findings become a focused, ranked plan - every opportunity scored by impact and effort, tied to the metric it moves, and backed by the deterministic signals beneath it.
- Opportunities ranked by impact × effort
- Each tied to the metric it's measured by
- Every move backed by the evidence beneath it
Strategy Engine
Your action plan
3 opportunities to focus on this month
Sample1Highest priority
Own the “best for everyday” recommendation
High impactMedium effortMeasured by ChatGPT recommendation share
- Why this
- You're named in the answer but rarely first, and never as the cited source - assistants lean on a retailer page instead of yours.
- What to do
- 1.Publish a comparison page targeting the “best everyday” buying question.
- 2.Add Product and FAQ structured data so assistants can ground on you.
- 3.Reclaim the citation with first-party proof points and reviews.
What you'll see: Move from mentioned to first-named on the everyday-use prompts.
Based on 4 signals
- 2
Turn recognition into top placement
Measured by Top-3 presence · 3 signals
High - 3
Win back the retailer citation layer
Measured by Owned citation share · 2 signals
Medium
Composed by the strategy engine from deterministic signals above. The wording is generated; the evidence is not.
Real attribution
Three layers of proof
“It worked” isn't a vibe. Because HeyOtis ingests your traffic and AI-bot logs, you can watch the crawlers fetch the fix, the assistants send the visitors, and the share move - with the evidence trail open.
The crawl
AI bots fetch the fix.
Within days of a move shipping, GPTBot, ClaudeBot and PerplexityBot fetch the changed pages - visible straight from your traffic and bot logs, because HeyOtis ingests them.
The visit
Assistants send the humans.
Referral sessions arrive from chatgpt.com and perplexity.ai onto the pages the move touched - real buyers, from real answers.
Referrals · 30-day window
AI referral sessions
Sample- chatgpt.com
- +128%
- perplexity.ai
- +64%
- gemini.google.com
- +41%
Sessions arriving from assistant surfaces onto the pages a move touched.
The lift
The metric moves.
Recommendation share is re-measured on the same prompts, before and after, with the evidence trail attached.
Attribution · 30-day window
ChatGPT recommendation share
SampleBefore1.4%
After4.9%
30-day window across tracked prompts, evidence trail attached
The compounding advantage
Every campaign makes the next one sharper
Proof isn't the end of the loop - it's the input to the next one. Every outcome, proven or disproven, reweights what the engine recommends next.
- Proven moves raise the weighting of moves like them
- Disproven moves get deprioritised - honestly
- Every cycle starts smarter than the last
The compounding loop
Recommendation share, by campaign cycle
- Cycle 1: 1.2% recommendation share - Baseline measured
- Cycle 2: 2.1% recommendation share - Comparison page proven
- Cycle 3: 3.4% recommendation share - Citation moves reweighted
- Cycle 4: 4.9% recommendation share - Schema moves prioritised
Built on evidence
It won't recommend what it can't prove
The cheapest thing to ship is a confident-sounding recommendation. The most expensive mistake is a confident-sounding wrong one. So the engine is built to refuse it.
Capabilities check
It only recommends a move it can verify got done. If a signal source isn't connected, it surfaces the gap instead of guessing.
Maturity gating
New signals ship dark, get measured, and only become client-visible once the data says they're trustworthy.
Validator-gated reporting
Every number traces back to source. The engine refuses to invent a statistic to make a point.
Platform + strategists
The engine finds the moves. Our strategists help you ship them.
HeyOtis pairs the Strategy Engine with hands-on GEO strategy. The platform does the analysis, the recommendations and the proof; our team helps you turn them into work that lands.
Monitored across
ChatGPT
Claude
Gemini
Perplexity
Google AI Overviews
Where this is heading
The operating system for brands in the age of AI search
The loop is the start. As assistants become how brands are discovered, served and transacted with, HeyOtis is building toward the full stack - see, serve, test and act on how AI represents you.
Get started
See the Strategy Engine
on your brand.
Book a 20-minute walkthrough. We'll run your brand, surface the highest-impact moves, and show you how the loop proves the lift.