Organise keywords for a cycle

Theory behind this: we target winnable SERPs and strategic fit rather than raw search volume. See Part 1 §1.4 SERP cliff and §1.5 Intent & cannibalization.

​

Inputs checklist (before you start)

• Granola transcript + enhanced notes from the monthly call
• Post-call notes, Slack messages, or docs from the client
• Previous month’s review docs (if not the first cycle)
• Access to Pi (memory already set up for this account)
• Any keyword lists already teased to the client (Notion proposal, strategy deck, etc.)
• (Optional) Keywords from Google Search Console

​

Step 1 — Feed Pi the call context

• Have Granola running during the call
• Export the transcript + enhanced notes afterward
• Collect post-call materials (Slack/email comments, client docs, previous reports)
• Paste everything into Pi with the prompt below
• Confirm Pi saved it to memory

here's the transcript and notes from the [month] review call with [client].
[paste Granola transcript + enhanced notes]

post-call notes:
[paste any additional client comments, corrections, or context]

attached: [any files]

save this to memory

​

Step 2 — Draft strategy + tag keywords

• Prompt Pi with the target keyword count + cycle constraints
• If you have a pre-existing list, include it and clarify these are candidates only (no landing pages created yet)
• Review Pi’s draft: does it match what the client said? any emphasized verticals missing? any keywords for products/features that aren’t live yet?
• Push back or add anything obvious before validation

based on the latest transcript and review, draft up a strategy.

suggest [N] keywords that we should tackle this cycle.
tag this cycle's keywords as '[cycle tag, e.g., apr to may 2026]'

[add constraints as relevant:]
- don't include keywords for [product/feature] — that's not live until [month].
  tag those for '[future cycle tag]' instead.
- focus on [verticals the client emphasized]
- [any other directives from the call]

here are keywords i've already proposed to the client as candidates for this cycle:
[paste list]

these haven't been used for landing pages yet — treat them as candidates
that still need validation. include them in your draft and add any others
you think we should target based on the call.

​

Step 3 — Validate, deduplicate, check for conflicts

• Send Pi the validation prompt (one prompt covers it all)
• Review the audit: removals, additions, recommendations for underperforming pages
• Flag any experimental angles you want to keep
• Approve, push back, or ask for alternatives on specific keywords
• (Optional) If the client wants a new angle, send the exploration prompt

can you use the keyword explorer and serp cliff to check through
all the keywords we've drafted for this cycle?

for each keyword:
- get US volume
- run serp cliff analysis
- check if the keyword already has a landing page
  (don't want duplicates or cannibalization)
- check if we already rank for it and at what position
- flag any keywords in our list that are too similar to each other
  (e.g., "ai chatbot tool" vs "ai chatbot software" — these would
  produce the same page, so pick the stronger one and drop the other)

also check any keywords tagged for next cycle and any untagged/unpublished
keywords in the dashboard — there might be opportunities to pull some
forward or better variants hiding in there.

then draft up a plan and reorganise the keywords.
retag if necessary but let's do it one step at a time.

before you start any actions, check with me to see if your strategy is sound

Note the last line: ask Pi to confirm its strategy before it acts. That “check with me before you start any actions” habit is standard across operators; it keeps you in control of destructive changes. See driving Pi well: 2.0.3.

note that [angle] is something [client] wants to target and test out.

use keyword explorer to expand on these experimental angles: [list angles]
and serp cliff to validate. use US volume to back the recommendations.

​

Step 4 — Approve and execute

• Tell Pi to execute the approved changes
• Confirm the execution summary matches what you agreed

looks good. go ahead and execute the changes.

​

Step 5 — Confirm final state

• Ask Pi for a final summary
• Run the final checks:
  • Keyword count is right (asked for 40, got 40)
  • Every vertical the client mentioned has keywords
  • Deferred keywords are tagged for the correct future cycle
  • No keyword overlaps an existing landing page
  • In the keyword content planner: no blogs/LPs already generated for these keywords, and no keywords too similar to each other (e.g. “ai translation” vs “ai translator”)
  • Strategy doc exists in /reports/

give me a summary of the final keyword state for [cycle tag].
include the total count, clusters, and flag anything still open.

​

Edge cases — quick lookup

​

Reference — why Pi recommends removing a keyword

​

Reference — why Pi recommends adding a keyword

​

Reference — the signal hierarchy

1. SERP cliff           — can we actually win this?
2. Strategic fit        — does the client want this?
3. Intent match         — right audience seeing our page?
4. No self-competition  — not competing with our own pages?
5. Volume               — how big is the prize?

Volume comes last, not first. That’s deliberate; see Part 1 §1.4.

​

Reference — keyword survival gates

​

Next / Related

• 3.2 Monthly planning flow: the planning beat this keyword cycle sits inside.
• 3.3 Weekly planning flow: where the tagged plan gets produced.
• Starting the first set of keywords: the first-cycle variant, seeded from the sales call.
• Part 1 §1.4 SERP cliff · §1.5 Intent & cannibalization: the theory behind winnable SERPs.