CClarivyBilingual AI Search Visibility

← 5-notebook set · Notebook 5 of 5 · 8-min read

Notebook 5 — Strategy

7 prioritized actions · effort estimates · expected citation-rate lift · This notebook answers: "What do we do this week, this month, and this quarter, and how much will each action move the needle?"

Data Provenance — Effort and lift estimates are PM estimates based on the 270 datapoints in this audit, cross-referenced with public guidance from the Princeton GEO paper, the llms.txt proposal, and Baidu Zhanzhang's official SEO documentation. License: CC-BY-4.0.

How to read this notebook

Each action below has 5 attributes: Action (what to do), Why (which finding in NB1–NB4 it addresses), Effort (PM-hours, single number), Expected lift (citation-rate delta on a re-run, ±0% stochastic noise), Owner (PM / Assistant / User). Actions are ordered by lift-per-effort, not by total lift.

The 7 actions

Action 1 — Publish the $99/$299/$1,499 pricing page in 3 places

Why: Notebook 2 §E shows that ChatGPT, Perplexity, and Gemini are explicitly naming Profound ($499/mo) and Otterly ($99/mo) price points on purchase-intent queries — and we are absent. The fastest way to get cited on a purchase-intent query is to have an indexable pricing page that the LLM can cite directly. The page already exists at clarivy.ai/#pricing, but it needs three additions: (a) a "Pricing as of 2026-06-11" dateline, (b) a machine-readable schema.org/Offer block, and (c) a separate /pricing/ URL (the current anchor-based pricing is not independently indexable).

Effort: 2 PM-hours (1h for the new page, 1h for the schema).

Expected lift: +1-2 mentions on the 5 purchase-intent prompts across the 3 Western engines. Estimated next-snapshot score: 2-3 / 45 (vs 0 / 45 today).

Owner: PM.

Deadline: 2026-06-13 (Day 4).

Action 2 — Pitch the "no SaaS does both" finding to 36kr + Search Engine Land

Why: Notebook 3 §2 shows that 36kr and Search Engine Land are the two highest-leverage publishers for the CN and Western engines respectively. Notebook 4 §"Quotable #1" shows that the gap statement — "no SaaS tool yet does both well" — is quotable, on-the-record, and attributed to Perplexity (a vendor the publishers will publish). Pitching the finding (not the product) is the cleanest path to a first citation.

Effort: 6 PM-hours (3h pitch writing + 3h back-and-forth on revisions).

Expected lift: +3-5 mentions on the methodology + competitive prompts across ChatGPT, Perplexity, Doubao, Kimi, MetaSo. A single 36kr post gets indexed into MetaSo's search index within 7-14 days; Perplexity picks up Search Engine Land within 3-7 days.

Owner: PM (drafts) + User (sends via admin@clarivy.ai to maintain the corporate voice).

Deadline: 2026-06-18 (Day 9, one week after the next snapshot baseline).

Action 3 — Publish 6 first-party methodology datapoints, one per NB2-C category

Why: Notebook 2 §C shows that methodology is the prompt category where the LLMs are most confident — and where there is no room to win the "what is GEO" answer. The opportunity is to be cited as a source for a methodology datapoint: e.g. "per Clarivy's Day-1 self-audit, 0/270 is the expected baseline for a brand-new domain" (we already have this one), or "per Clarivy's 2026 audit of 12跨境 brands, the average mention rate on Perplexity is 14%". Each datapoint must have named authorship, CC-BY-4.0 license, and a stable URL. Six posts, one per methodology sub-category, with verifiable data — that's the durable lever.

Effort: 12 PM-hours (2h per post, including data + writing + schema).

Expected lift: +4-6 mentions on the methodology prompts. The pattern from the Princeton GEO paper data and the llms.txt proposal data is that LLMs begin citing a new source within 30-60 days of consistent publishing. We expect to see lift on the 2026-08-01 snapshot, not the 2026-07-01 one.

Owner: PM.

Deadline: 2026-07-01 (Day 20, aligned with the next snapshot date).

Action 4 — Submit 5 Chinese-engine crawl registrations (Baidu Zhanzhang + 4 more)

Why: Notebook 3 §2 shows that Baidu Zhanzhang, ByteDance Juliang, Shenma, Sogou, and Bing are the upstream crawlers that feed the CN engines. Submitting clarivy.ai to all 5 is a hygiene step, not a magic lever — but it is the prerequisite for the CN engines to ever see us. Doubao, Kimi, ERNIE, and MetaSo all rely on these crawlers (or on Bing/Google for cross-language content). Submission is free; the only cost is the time to fill the forms.

Effort: 2 Assistant-hours (forms are tedious but low-judgment).

Expected lift: +0 mentions on the 2026-07-01 snapshot, +1-2 mentions on the 2026-08-01 snapshot. Crawl → index → LLM-cited latency is 30-45 days on the CN engines.

Owner: Assistant.

Deadline: 2026-06-15 (Day 5).

Action 5 — Install /llms.txt at the root of clarivy.ai

Why: Notebook 2 §C and Notebook 4 §"Quotable #3" both show that the llms.txt proposal is supported by 3 of the 9 engines (ChatGPT, Perplexity, Claude) and is being recommended as a methodology best-practice by 4 more (Kimi, ERNIE, DeepSeek, Doubao). The proposal is at llmstxt.org. Installing the file is a 30-minute task. We won't see a lift from this in the 30-day window, but the 90-day lift is meaningful.

Effort: 0.5 PM-hours (write the file, deploy).

Expected lift: +0 mentions in 30 days, +2-4 mentions in 90 days (Claude + Perplexity in particular).

Owner: PM.

Deadline: 2026-06-12 (Day 2, today).

Action 6 — Get a Wikidata QID + a Wikipedia "notability" draft

Why: Notebook 3 §2 shows that Wikipedia + Wikidata are the two upstream sources that the LLMs cite most consistently for entity recognition. A Wikidata QID for HG-Solution Co., Limited (CR 80121024) is straightforward — the company is real, registered, and has a public CR number. A Wikipedia draft is harder (notability threshold is high), but a stub article with the CR number and the methodology page is enough to start. The lift is small but the duration is permanent.

Effort: 4 PM-hours (Wikidata: 1h; Wikipedia draft: 3h for the WP:NPOV dance).

Expected lift: +1-2 mentions on the brand-specific prompts across all 9 engines, starting on the 2026-08-01 snapshot. Permanent once live.

Owner: PM (Wikidata) + User (Wikipedia draft, because it requires an account under the user's real name for WP:NPOV).

Deadline: 2026-06-30 (Day 19).

Action 7 — Publish the methodology page in zh-Hans

Why: Notebook 2 §A and §B show that the Chinese engines (Doubao, Kimi, ERNIE) are surfacing Chinese vendors for buying-intent and competitive prompts — and we are absent. The methodology page is the single highest-quality English-language content on clarivy.ai, and a 1:1 zh-Hans translation (not a machine translation — a hand-edit by the PM, with the same Named-Entity markup) will be the single most-leveraged Chinese-language publishing action. CN engines index Chinese content faster than English content; the expected lift is asymmetric.

Effort: 3 PM-hours (translate, deploy, add hreflang).

Expected lift: +2-4 mentions on the 5 Chinese engines' buying-intent and competitive prompts, starting on the 2026-07-15 snapshot.

Owner: PM.

Deadline: 2026-06-20 (Day 10).

Effort and lift summary

#ActionEffort30-day lift90-day liftOwnerDeadline
1Pricing page in 3 places2 PMh+1-2+2-3PM2026-06-13
2Pitch 36kr + Search Engine Land6 PMh+3-5+5-8PM+User2026-06-18
36 first-party methodology datapoints12 PMh+0+4-6PM2026-07-01
45 CN-engine crawl submissions2 Ah+0+1-2Assistant2026-06-15
5/llms.txt at root0.5 PMh+0+2-4PM2026-06-12
6Wikidata QID + Wikipedia stub4 PMh + User+0+1-2PM+User2026-06-30
7Methodology page in zh-Hans3 PMh+1-2+2-4PM2026-06-20
Total27.5 PMh + 2 Ah+5-9+17-31

What the next snapshot will look like

At the current action-execution rate, the projected self-audit #2 (2026-07-01) score is 2-4 / 270 mentions, with the lift concentrated in the purchase-intent and competitive categories (because Actions 1, 2, 4 are the ones with the shortest latency). The methodology category is not expected to move until 2026-08-01 at the earliest, because the LLMs are slow to incorporate new sources. The brand-specific category is the last to move — we expect 1-2 brand-specific mentions on the 2026-08-01 snapshot, not before.

What we will NOT do

From the v3 plan §"Day 1 honest conclusions", restated for the next 30 days:

  1. No link buying. Buying links is a Google penalty since 2012, and the LLM equivalents (citation networks, entity associations) are even more expensive to fake.
  2. No PBNs. Same reasoning.
  3. No fabricated llms.txt statistics. Every datapoint in the llms.txt file will be either (a) a real datapoint from this audit, or (b) a real datapoint from a future customer audit (with the customer's written consent).
  4. No "as seen in" logos that aren't true. The Day-1 self-audit shows that 0/270 is the baseline. We will not add "as cited in ChatGPT" or "featured by Perplexity" until those citations actually exist.

How we will measure the next snapshot

Self-audit #2 is on 2026-07-01. The 5 notebooks will be re-run end-to-end. The new score will be published at clarivy.ai/audit/self-audit-02/ on the same day. The methodology will be identical to this audit (same 30 prompts, same 9 engines, same binary mention-count, same data-provenance footer). The only deltas will be: (a) a position-weighted scoring column added to NB1, (b) a freshness and sentiment column added to NB3, and (c) the action-by-action lift tracking added to NB5.

← Back to 5-notebook set   1-page summary   Download raw JSON