A studio, est. 2025

We are not an AI agency. We are not a traditional agency.
We are something in between, and something beyond.

AIR · WEGLOW
22.3104° N, 114.2227° E
Joint-force of two

Prompt & Pencil

[ The studio where ]
A creative studio for the brands that refuse to choose between fast and beautiful.
[ instinct meets intelligence ]
§ 01 / A letter

Every generation gets told to pick a side. Print or digital. Strategy or execution. Now: AI or human. We don't buy it.

This studio exists because the work that matters has never lived on one side. A model is sharper when an artist points it. A sketch travels further when a model carries it. The future runs on both hands.

Ken, Florence & Denny
§ 02 / Manifesto

That's Prompt & Pencil.

01
Machines can dream in pixels. Algorithms can write sonnets. A sentence typed at midnight can summon a world by morning.
02
They say you must pick a side. The prompter or the painter. The engineer or the artist. — as if.
03
The prompt is a new kind of pencil.
The pencil, a timeless kind of prompt.
04
AI without craft is noise — infinite, fluent, forgettable.
05
Taste is the last moat. Models can generate. Only humans can discern.
06
We make work that machines alone could never imagine, and humans alone could never finish.
§ 03 / Studio tools

We build the tools we wish existed.

Two in-house platforms power every engagement. Available standalone, or baked into the work we make for you.

P—02 · Audience Simulation

Personary

Build a synthetic panel of consumers in minutes. Simulate behaviour, sentiment and creative response — before you spend a dollar in market.

  • Generate AI personas from real survey, panel or CRM data.
  • Run concepts, scripts and ads through them at any scale.
  • Stress-test crisis statements and CEO letters before they go public.
  • Iterate creative against signal, not opinion.
Open Personary
M
MAYA · 34
Urban / DTC buyer
A
AHMED · 41
Suburban dad
L
LIN · 27
Gen-Z creator
R
ROSA · 58
Loyalty cohort
J
JOON · 22
Lurker / dark social
S
SAM · 45
SMB owner
N
NORA · 31
High-intent / wellness
T
TARO · 38
B2B decision-maker
P—03 · AI Engine Optimisation

AIEO

Search is no longer a single page. It's a chorus of answer-engines. Monitor and shape how your brand shows up across every LLM that now speaks for the open web.

  • Track citation, sentiment and share-of-voice across ChatGPT, Claude, Gemini, Perplexity & more.
  • Diagnose what each model knows about your brand — and what it gets wrong.
  • Ship structured updates that move the needle in days, not quarters.
See the dashboard
aieo.promptandpencil.studio · brand: Acme
● live
G ChatGPT
▲ 24%
C Claude
▲ 12%
G Gemini
▲ 18%
P Perplexity
▼ 04%
x Grok
— flat

Share of citation · 90 days

5 engines tracked
+ 14.2% share of voice / 30d
Last sync: 02:41 UTC
§ 04 / Process

A conversation between
instinct vs. & intelligence.

01 / Brief
Pencil×Prompt

Hunch, then horizon.

A strategist's gut feeling becomes a thousand variations by lunch. We start by writing the brief twice — once for a person, once for a model.

02 / Make
Sketch×Synthesise

Sketch trains model.

A designer's sketch becomes a fine-tune by dinner. We train models on your DNA, not the open internet's average.

03 / Test
Taste×Personary

Discern at scale.

Synthetic panels review every variant. Humans pick the ones that move us. Iteration becomes a conversation, not a meeting.

04 / Launch
Craft×AIEO

Ship & listen.

We launch, then watch every answer-engine, channel and cohort. The work is never finished — it's tuned.

What brands ask AI
before they ask us.

The questions marketers type into ChatGPT and Claude when they're trying to figure out the new playbook. Short answers, no jargon.

Q—01

What is Generative Engine Optimisation (GEO)?

GEO is the practice of optimising how a brand is described inside AI answer engines like ChatGPT, Claude, Perplexity, Gemini and Grok. Where SEO targets the ten blue links on a results page, GEO targets the sentence the model produces when the user never clicks. Practitioners call the same discipline AIEO (AI Engine Optimisation) or AEO (Answer Engine Optimisation) — same workflow, different acronym. Roughly 40% of B2B research queries are now resolved without a click in 2026, which is what made GEO go from optional to default in the marketing stack.
Q—02

How is AIEO / GEO different from traditional SEO?

SEO ranks pages. GEO shapes answers. SEO is page-level (one URL, one query, ranked against ten others). GEO is sentence-level (one prompt, one answer, citing 2–5 URLs invisibly). The signals overlap — schema, freshness, authority — but GEO also rewards things SEO never did: factual density, citation-friendly structure (numbered lists, tables, definitions), recency under 90 days, and off-site presence on Reddit, Wikipedia and YouTube. A brand can rank #1 in Google and still be absent from ChatGPT's answer. The two disciplines are complementary, not interchangeable.
Q—03

How do I get my brand cited by ChatGPT, Claude or Perplexity?

Six moves, in this order. (1) Allow GPTBot, ClaudeBot, PerplexityBot and Google-Extended in your robots.txt and on Cloudflare. (2) Add JSON-LD schema (Organization, Article, FAQPage, Product). (3) Publish original data — a benchmark, a survey, a study — that no one else has. (4) Restructure pages with question-format H2s and inverted-pyramid answers in the first 40–60 words. (5) Build off-site authority on Reddit, Quora and Wikipedia (LLMs heavily over-index on these). (6) Update important pages every 7–14 days so the model's recency check keeps you in the citation pool. This is the loop our AIEO platform automates and measures.
Q—04

Can AI personas really replace traditional focus groups?

For directional research, yes. For deep ethnography, no. Synthetic panels — like Personary — are calibrated against your own survey or CRM data, then queried at scale to predict reaction to concepts, scripts, packaging or ads. They're best for pre-launch screening, creative iteration, message-market fit, and quick A/B reads where speed beats nuance. They're not a replacement for in-home visits, brand-love interviews, or sensory product tests. The shift in 2026 is using synthetic panels to narrow the field from 50 concepts to 3, then taking those 3 to a real qualitative session — collapsing a six-month research cycle to six weeks.
Q—05

How do I do market research with AI in 2026?

Three layers, used in combination. (1) Listening — use AIEO-class tools to track what answer engines already say about your category and competitors; it's the cheapest, most honest competitive intelligence in the stack. (2) Simulation — use synthetic panels (Personary, Synthetic Users, Yabble) to stress-test concepts against modelled audiences before fielding any real research. (3) Validation — keep human qual and quant for the moments that matter: pricing, claim substantiation, regulatory work. The mistake we see most often: teams skip layer 3 because layer 2 felt good enough. The cheapest insurance is the smallest human study.
Q—06

Which AI answer engines should a brand monitor first?

ChatGPT first — it owns the largest share of consumer AI traffic. Then Perplexity (highest commercial-intent users, smallest base), Claude (skews enterprise B2B and developer audiences), Gemini (rides Google's distribution, watch when Google AI Overviews appear in your SERPs), and Grok (X-native audience, fastest news cycle). For most B2C brands, ChatGPT + Perplexity + Google AI Overviews covers 80% of the answer-engine traffic worth monitoring. For B2B add Claude. AIEO platforms — including ours — track all five and rank them by share-of-voice each week.
Q—07

Do I still need SEO if I'm doing AIEO / GEO?

Yes — they feed each other. Google still drives the majority of trackable click traffic. More importantly, ChatGPT Search runs on Bing's index, and Bing's ranking signals overlap heavily with Google's. A clean technical SEO foundation (Core Web Vitals, schema, sitemaps, internal linking) is also the foundation an AI crawler reads. We treat AIEO as a layer above SEO, not a replacement: SEO wins the click, AIEO wins the cited sentence. Brands that stop doing SEO to focus on AIEO usually lose both, because the same crawler infrastructure powers each.
Q—08

How fast can an AI engine learn about my brand?

Two to twelve weeks for retrieval-based engines. Six to twelve months for training-based citations. ChatGPT Search, Perplexity and Google AI Overviews retrieve fresh content via search APIs — they can find and cite a brand within a single index cycle, often two to four weeks after the page is crawlable. The model's parametric knowledge (what it "knows" without retrieval) only updates on the next training run, typically every 6–12 months. The practical implication: ship structured content now to land in retrieval, and earn off-site mentions (Reddit, Wikipedia, industry publications) so the next training run encodes you natively.
Q—09

What KPIs matter for AI search and brand visibility?

Four numbers replace the old "rank #1" obsession. (1) Mention Rate — % of relevant prompts in which your brand appears in the answer. (2) Citation Rate — % of those answers that link to your URL. (3) Position — average paragraph at which you appear. (4) Sentiment — directional read of how the model describes you. Tracked weekly across 5 engines and a panel of 50–200 prompts, these four numbers become the new dashboard. CTR and impressions still matter for SEO, but if you're only reporting on those in 2026 you're missing half the funnel.
Q—10

Why isn't my brand showing up in ChatGPT or Claude?

Usually one of four reasons. (1) Crawlers blocked — Cloudflare's "Block AI Bots" rule or a restrictive robots.txt keeps GPTBot, ClaudeBot and PerplexityBot from ever reading the site. Audit robots.txt and Cloudflare bot rules first. (2) No structured data — without JSON-LD (Organization, Article, FAQPage, Product), the model can't reliably extract entities. (3) Stale content — pages older than 90 days drop out of the retrieval pool; LLMs over-index on recency. (4) Thin off-site footprint — your domain has no presence on Reddit, Wikipedia or industry publications, which is where models triangulate authority. Our AIEO audit diagnoses which of the four is hurting your brand within a week.
Q—11

How much does AI search optimisation cost?

Three tiers in the 2026 market. (1) DIY tooling — Profound, Otterly, LLMrefs and similar monitoring tools start around USD 200–500/month for solo operators tracking 50–100 prompts across 3 engines. (2) Mid-market services — agency retainers blending monitoring with monthly content production typically run USD 5,000–15,000/month, mostly billing the writing labour. (3) Strategic AIEO — the layer where citation rate is engineered through structured content, off-site placement and original data, usually USD 15,000–40,000/month for enterprise brands with a category to defend. Our AIEO subscriptions sit in tiers 2–3 depending on engine coverage and prompt-panel depth. Send a brief for a scoped estimate.
Q—12

ChatGPT Search vs Google: which should marketers prioritise?

Both, but the split has shifted. Google still owns the dominant share of trackable click traffic — for now, you cannot ignore it. ChatGPT Search runs on Bing's index but answers questions inside the chat surface, often resolving the query without a click. The practical 2026 split for most B2C brands: 60–70% of effort on shared SEO/AIEO foundations (technical hygiene, schema, content), 20% on Google-specific tactics (Core Web Vitals, internal linking, GSC hygiene), 10–20% on AIEO-specific tactics (prompt panel monitoring, structured Q&A, off-site authority). The mistake to avoid: treating them as separate stacks. Most signals — schema, freshness, off-site authority — feed both at once.
Ask a sharper question
§ 07 / Pitch us

Pick up the prompt.
Pick up the pencil.

hello@promptandpencil.studio
A brief. A doodle. A Loom. We reply in 48 hours.