Measurement for what AI assistants recommend.
Recall asks the engines the questions your buyers ask, keeps every answer and source, and shows why you are on the list or missing from it. In intervals, not scores.
Vector databases · who the engines name
Recorded run · 17 Sep 2026 · n = 24
A real gap
83% to 21% falls outside both intervals.
Overlapping intervals
The sample cannot rank these four, so Recall does not.
A real gap
83% to 21% falls outside both intervals.
A different cause
pgvector is an extension and the questions ask for platforms. Low visibility is not the problem.
01What a figure carries
A score is a claim. A Recall figure shows its working.
A visibility score
88
No sample size, no range, no answers behind it. Higher than 83, so ranked above it.
A Recall figure
88%
- Interval
- 69–96%
- Sample
- 24 answers to unbranded buyer questions
- Ranking
- Not ranked: 3 rivals overlap
- Evidence
- Every answer and cited page, stored
01
Evidence
Every figure opens onto the answers and pages behind it, captured when they were cited. A source rewritten later is flagged as changed.
02
Diagnosis
Absence has four causes: unreachable, unmentioned, unrecognised, or passed over. Each gets a different fix.
03
Proof
The baseline is locked when work ships. A lift is called only when the intervals separate.
04
Revenue
Sessions and conversions that arrive from AI assistants, either side of a shipped change.
ChatGPT, Gemini, Perplexity, Claude, Copilot, Grok, DeepSeek, Mistral, and Google’s AI Overviews and AI Mode, through the providers you configure.
All capabilities02Who it is for
Two questions: are you on the list, and is it true?
- Venture funds and portfoliosMeasure the fund's own founder-facing visibility, then give portfolio teams the same evidence-backed operating loop.One relationship can improve an entire portfolio.
- Technical categoriesDevelopers and platform teams genuinely shortlist this way. We measured it: every probe in AI infrastructure returned a usable answer.Named in the comparison, or not.
- First-timers and newcomersA family choosing a school, someone who just moved cities, parents buying gear they have never owned. Nobody to ask, so the whole search happens in the open.Win the questions asked before a tour, call, or purchase.
- Procurement-led and regulatedYour buyers use vendor lists, not prompts. Journalists, investors and engineers ask anyway — and an engine with thin evidence answers from memory rather than declining.What was said, by which engine, sourced or not.
03What runs without you
The problem is not the work. It is that nobody is watching.
- Daily
The questions your buyers ask, put to every engine again
Answers drift. A shortlist you were on in March is assembled from different sources in September, and nothing announces the change.
- Daily
Every page that cited you, fetched again and compared
The article that earned you a recommendation gets rewritten or deleted. The citation quietly stops supporting the claim, your visibility erodes, and the cause is invisible.
- Every 6 hours
Shipped changes read against their locked baseline
Work gets called a success because the number moved, or abandoned because it did not move yet. Both happen when nobody is holding the before.
- Hourly
A competitor appearing where you used to be named
You find out from a lost deal months later, at which point the sources that made the case for them are established.
- Continuous
The measurement itself, checked for having stopped
A dashboard keeps rendering the last result. The numbers look plausible and nobody discovers the data stopped until they compare it to reality.
Each reaches you only when it has something to say. A weekly note reporting no change teaches you to filter it, and takes the urgent one with it.
04Pricing
Start with a measured pilot. Expand when the evidence says so.
01Shortlist Pilot
€1,000
credited in full if you continue
Four weeks. One brand, one market, and an answer you can act on.
- Buyer-question workshop and approved scope
- Multi-engine baseline across ChatGPT, Perplexity and Gemini
- Competitor and citation evidence, with the stored answers behind it
- Why you are missing each question, not just that you are
- Ranked actions with paste-ready deliverables
- Before-and-after measurement design and the first read
02Managed Recall
Custom
ongoing program
Keep the shortlist, evidence, and action queue current.
- Everything in Shortlist Pilot
- Scheduled prompt monitoring
- Configurable engine coverage
- New and dropped source alerts
- Evidence-backed action queue
- Implementation impact tracking
03Portfolio & Partner
Custom
Monitor a fund, its portfolio, or multiple client brands.
- Everything in Managed Recall
- A separate isolated workspace per brand
- Custom branding and domain
- SSO and role controls
- Pilot design and priority support
- Roadmap: fund-level reporting across a portfolio
00Free first check
Before you buy a measurement, we check whether the engines can reach your site, tell who you are, and find enough to retrieve. If none of those is the problem, no on-site work will move it, and we will tell you so.
05FAQ
Frequently asked questions
What is Recall?
Recall measures what AI answer engines say about a company. It runs real buyer questions through configured engines on a schedule, records who is recommended and which sources are cited, keeps the stored answer behind every figure, and turns what it finds into changes a team can ship and then measure. Where buyers shortlist with AI, that means share of answers. Where they do not, it means what the engines assert about you and how much of it has no source.
What is Answer Engine Optimization (AEO)?
Answer Engine Optimization (AEO) — also called Generative Engine Optimization (GEO) — is the practice of measuring and improving whether AI answer engines accurately recommend and cite your brand. Instead of ranking in a list of links, the goal is to earn a credible place in the answer and the shortlist.
Who is Recall for?
Recall is strongest where a buyer has nobody to ask. Referral beats search whenever someone can call a friend who has already made this choice, so AI research dominates for people new to a place, doing something for the first time, buying across a border, or working in a field that moves faster than anyone's experience of it. That covers venture funds and their portfolios, AI infrastructure, Web3, coworking, private schools and kindergartens. Procurement-led sectors such as defence and aerospace are served differently: nobody shortlists a supplier from an AI answer there, so the measurement is accuracy rather than position.
Which AI models does Recall track?
Recall supports configurable coverage for engines including ChatGPT, Google AI Overviews and AI Mode, Perplexity, Gemini, Claude, Copilot, Grok, Mistral, and DeepSeek. The engines available in a deployment depend on the providers and credentials you configure.
How is Recall delivered?
As a managed platform that we run. Each brand's data is isolated at the database level, every answer and citation exports in full whenever you ask, and the model coverage, spend limits, access controls, and operating support are agreed as part of the pilot. There is no self-hosted edition today.
Is Recall an SEO or content agency?
No. Recall is the measurement and decision layer: it shows who AI recommends, preserves the evidence, and ranks defensible actions. Your team or delivery partner can implement those actions, and Recall measures whether the answers changed afterward.
Start with the shortlist that matters most.
One brand, one market, and the buyer questions closest to revenue or deal flow. Four weeks to a baseline, the first changes, and a read on whether they moved anything.