From 'monitor everything' to 'monitor by Focus': one archetype per Focus

Here’s the conclusion up front: monitoring all of your AI visibility “uniformly, everywhere” is both a waste of money and the single biggest reason your decisions stay murky. Piling undifferentiated citation counts onto a dashboard does not answer the one question that matters — “so what do we actually fix?” What you need instead is to monitor by Focus: before you collect a single data point, you fix the kind of question you’re answering. And in VeReach GEO, a Focus isn’t a loose “track the brand” label — it monitors exactly one archetype, one of eight query-form classes, each of which comes with a coherent KPI family and a brand guardrail that keeps the query set honest.

This article explains why uniform monitoring breaks down, lays out the eight archetypes in a single table, and gets practical about how to scope your first two or three focuses.

---

1. Why “monitor everything uniformly” breaks down

The reflex of most teams new to AI-search monitoring is “run every major engine, on every query we can think of, every day.” It looks thorough and safe. It is wrong on two counts.

First, cross-engine overlap is small. The URLs ChatGPT cites and the URLs Perplexity cites are largely disjoint, and AI citations themselves swing 40–60% month-over-month. The implicit premise — “watch all the engines uniformly and you’ll see the whole picture” — simply doesn’t hold. Unless you pick what matters, every engine’s number is just a half-baked sample of something you never scoped.

Second, different questions need different metrics. “Are we recommended in our category?” and “When AI states our price, is it correct?” are not the same monitoring job. They want different queries, a different success metric, and a different definition of “good.” Painting both with one playbook — a single brand keyword fired at every engine and a generic “mentions” count — is GEO’s most common, and most expensive, waste. The market itself is moving too fast to waste budget this way: AI search in Japan grew 3.5× in eight months (Hakuhodo), and 46.4% of Japanese B2B buyers report finding a vendor through AI (LANY).

Monitoring is a cost. Observing AI surfaces under real-world conditions requires independent environments and sampling across region, language, device, and personalization. That is exactly why deciding what kind of question each Focus answers is as strategic as deciding what to track at all.

---

2. What a Focus actually is

The hierarchy in VeReach GEO is simple: Organization → Project → Focus → KPIs. A Project is a pure UI folder — it carries no schema, it just groups things. The unit that does the real work is the Focus: the monitored entity that turns “watching numbers” into “answering a business question.”

A Focus operationalizes what you track × the business question you’re answering — but the concrete, shipped structure is precise. The governing rule:

A Focus monitors exactly one archetype. One archetype yields a homogeneous query set, which is fairly explained by a single KPI family. That’s the whole reason a Focus stays interpretable instead of becoming another dashboard of unrelated numbers.

If you try to cram “are we recommended in our category” and “is our stated price correct” into one Focus, the query set goes heterogeneous and no single KPI family can fairly score it. So you don’t — you split them into two focuses, each on its own archetype.

Each Focus also owns the machinery to run that archetype well:

• brandName + typed brandVariants (katakana / hiragana / kanji / romaji / ascii — the spelling drift specific to the Japanese market)
• trackedEntities with a role (self / competitor / reference / other) — this feeds Share of Voice and the competitor gap
• a prompt set with samplesPerCell (default 10 — each prompt × engine is queried N times to average out LLM randomness)
• a web-grounding mode (all / JP tier-1 only / allowlist / blocklist)
• a cadence for scheduled auto-collection
• its own time-series store — every run becomes one point on that Focus’s timeline

Engines tracked are ChatGPT, Gemini, Claude, and Perplexity (via OpenRouter).

---

3. The eight archetypes

This is the centerpiece. Every Focus picks exactly one of these eight query-form classes. The brand guardrail — two axes, self and competitor, each forbidden / optional / required — is what keeps the query set on-form: a brand-named comparison can never leak into a brandless category-recommendation Focus.

#	Archetype	The GEO question it answers	Brand guardrail
1	Category recommendation	”When users ask AI for options in my category, am I recommended / ranked / first?“	brandless — no brand named
2	Head-to-head comparison	”In a named X-vs-Y comparison, does AI favor us, and are the claims accurate?“	self + competitor both named
3	Brand-direct reputation	”Asked directly about us, is the portrayal accurate, positive, crisis-free?“	self only
4	Factual accuracy	”When AI states our price / specs / features, are they correct vs official?“	self only
5	Campaign / freshness	”Does AI know our latest launch / campaign and cite the official source?“	self only
6	Authority / citation	”On a topic we own the data for, does AI attribute and cite us?“	brandless — names the topic
7	Differentiation salience	”Are our differentiators present, or are we flattened to ‘same as the competitor’?“	self + competitor
8	Intent coverage	”Across brandless category intents, which do competitors monopolize, which do we miss?“	brandless

One nuance worth keeping straight: the subject angle — whether you’re framing around the company, the brand, or a specific product — is orthogonal to the archetype. It only fills the {{brand}} slot in prompts; it does not change which archetype you’re on. The guardrail governs the form; the subject angle decorates it.

---

4. KPIs: eight builtins plus the archetype’s custom family

Every Focus, regardless of archetype, tracks the same eight builtin KPIs, computed from each run’s score summary and projected onto the time-series as one point per run:

1. Visibility — Visibility Score VS = 0.4·Coverage + 0.3·Position + 0.3·Influence
2. Share of Voice
3. Weighted share (Japanese-source-weighted)
4. Sentiment ratio
5. Own mentions
6. Competitor mentions
7. Citations
8. Competitor gap

That’s the common spine. On top of it, each archetype suggests its own custom KPIs — LLM- or operator-defined extraction dimensions (valueType: a label, a score, a percent, or a presence flag) that the authoring agent materializes into the Focus’s response schema. The point is that the custom KPI must fairly explain the queries that archetype produces. Two examples:

• Head-to-head comparison → “conclusion favors us / competitor / tie” (label), “claim accuracy vs official” (score).
• Factual accuracy → “stated fact correct / partially / incorrect / not-stated” (label), “hallucinated detail present?” (presence).

Don’t dump unrelated numbers on a dashboard — produce a diagnosis that points straight to action. Because the archetype fixes the query form and the KPI family in one move, “what to track” and “how to score it” stay aligned by construction, not by hope.

A Focus can also be authored by a conversational agent: describe the brand, pick the archetype, and the agent proposes the prompt set, the custom KPIs, and the response schema — then triggers a run. Designing a Focus becomes a dialogue.

---

5. How to scope your first two or three focuses

Once you’ve decided to stop “monitoring everything,” build at most two or three focuses and expand only after they pay off. Pick archetypes, not keywords.

Focus 1 — Category recommendation (brandless)

Start with archetype #1. Ask the question every category leader needs answered: when a user asks AI for options in your category without naming you, are you recommended, ranked, and ideally first? This is brandless by guardrail, so the query set stays a clean test of category presence rather than a vanity check on your own name. Visibility, Share of Voice, and competitor gap carry most of the signal here.

Focus 2 — Head-to-head comparison (self + competitor)

Next, if you can name the rival you keep losing to, run archetype #2. Both brands are required by the guardrail, so every query is a genuine X-vs-Y comparison. The custom KPIs — “conclusion favors us / competitor / tie” and “claim accuracy vs official” — turn a fuzzy worry into a tracked number you can move.

Focus 3 (optional) — Factual accuracy or Brand-direct reputation (self only)

The third depends on your risk surface. If AI keeps misstating your pricing or specs, run archetype #4 (Factual accuracy) and watch the “stated fact correct / incorrect” label plus the hallucination-presence flag. If you’re worried about how the brand at large is portrayed, run archetype #3 (Brand-direct reputation) and lean on sentiment ratio.

Priority	Archetype	Brand guardrail	Why do it first
1	Category recommendation	brandless	Tests true category presence; competitors are already eating this surface
2	Head-to-head comparison	self + competitor	Turns “we keep losing to X” into a tracked, movable number
3	Factual accuracy / Brand-direct reputation	self only	Chosen by your risk surface — wrong facts or a reputation worry

Run those two or three, each on its own archetype and its own time-series. That alone turns monitoring from “a pile of floating numbers” into “a diagnosis tied to the business.”

---

6. In summary

AI-visibility monitoring is won on focus, not coverage. Cross-engine citation overlap is small, AI citations swing 40–60% month over month, and different business questions demand different metrics — these facts rule out “uniformly, everywhere” from the start.

So monitor by Focus, and remember the governing rule: one Focus, one archetype, one KPI family, one brand guardrail. Eight builtin KPIs give every Focus a common spine; the archetype’s custom KPIs make the diagnosis answer the actual question. VeReach GEO operationalizes that whole chain as a product unit, lifting “watching rankings” into “watching the business.” Start with just two — category recommendation and your sharpest head-to-head.

If you want targeted monitoring and a diagnostic report under real-world conditions, scoped to the right archetypes for your business, get in touch. For product details, see the VeReach GEO solution.

---

A note on methodology: the archetype model, KPI families, and the Visibility Score formula reflect VeReach GEO’s shipped product as of June 2026; the AI-citation volatility, Japan AI-search growth, and JP B2B buyer figures come from public reporting (Hakuhodo, LANY) as of the same date. AI surfaces evolve fast — verify against the latest official guidance before any key decision.