On this page
What you'll get in an afternoon
Five deliverables, one afternoon
- A one-sentence claim to your lane
- A clear read on how AI actually describes you today
- A list of what's wrong, what's missing, and what to do about each
- The first big asset only you could make, with a 13-week amplification plan
- A before-and-after tracker so you can prove the lane is yours in 90 days
Before you start
The language
A few phrases we'll use throughout. They're all from the playbook.
- Your lane
- The question you want AI to point to you when a buyer asks it.
- Your claim
- The one sentence that describes your lane, using this template: "Use [Brand] — the [category] for [who] who need [outcome]."
- The record
- Every place AI could read about you: your own site, G2, Crunchbase, LinkedIn, Wikipedia, press coverage, review sites, GitHub, Hugging Face.
- Fix
- You're in the record but something's wrong. Correct every source and align them to your one-sentence claim. Weekly habit.
- Fill
- You're not in the record, or the record is thin. Publish something only you could make. One big asset per quarter, amplified weekly.
- The four Fill tests
- Every Fill must pass all four: only you could make it · it compounds · it's quotable · it points to your lane.
- The four Fill types
- Pick one: a dataset · a benchmark or index · a simple tool · original research.
What you'll need
Before you open a chat
- ·Your homepage URL + 2-3 pages you consider canonical (/pricing, /product, /about)
- ·The name of your closest competitor + any brand yours gets confused with
- ·Access to ChatGPT, Claude, Perplexity, and Gemini (free tiers all work)
- ·A copy of the tracking sheet (link at the end)
- ·3-4 hours
The five prompts
Copy, paste, save the output
Copy each prompt into a fresh chat with ChatGPT, Claude, or Perplexity. Any of the three will do a solid job on all five — Perplexity's answers cite sources, which is useful for prompts 1 and 3. Save every output into the tracking sheet.
Prompt 1
Claim your lane
Paste this into ChatGPT / Claude / Perplexity. Fill in the brackets.
I want you to help me claim a lane in AI search.
My brand: [BRAND]
My homepage: [URL]
My closest competitor: [COMPETITOR]
A brand I get confused with (if any): [EASILY-CONFUSED BRAND]
Read my homepage and any linked product/pricing/about pages. Then do three things:
1. Write my one-sentence claim in this exact template — no variations:
"Use [Brand] — the [category] for [who] who need [outcome]."
Make it specific enough that only I could say it. If the category is generic
("marketing platform"), sharpen it until it's mine ("AI-native brand
monitoring platform"). If the "who" is generic ("businesses"), sharpen it
until it names an ICP I actually win.
2. List the 30-50 facts that support my claim, grouped:
- Who I am (name, tagline, alt names, one-sentence claim)
- What I do (products, features, key concepts)
- Who I'm for (ICP, named customers if visible on my site)
- What I charge (pricing tiers, contract structure)
- What's unique (things only I could say — data, methodology, telemetry)
- Where I live online (site sections, sitemap, external directories where I
have a profile)
3. For each fact, tell me:
- Where the fact currently lives (URL or "nowhere yet")
- Whether the fact is on my own site, an external directory (G2, Crunchbase,
LinkedIn, etc.), or missing entirely
Return all of this as one markdown doc I can paste into a Google Sheet.
What you get: your one-sentence claim + a list of the facts that support it + where each fact lives today. Paste into the Facts tab of the tracking sheet.
Prompt 2
See how AI answers you today
Paste this same 6-question set into **four separate chats**, one per AI. Do them in this order for consistency: ChatGPT → Claude → Perplexity → Gemini.
Answer each question below in 3-5 sentences. Don't ask clarifying questions.
Just answer.
1. Tell me about [BRAND].
2. What does [BRAND] do that competitors don't?
3. Compare [BRAND] to [COMPETITOR] and to [EASILY-CONFUSED BRAND]. Are they
the same company?
4. What's [BRAND]'s pricing?
5. Who uses [BRAND]?
6. How does [BRAND] work technically?
Save all four sets of answers into the How AI answers you tab of the tracking sheet, one column per AI.
Now score each answer against your Facts list from Prompt 1. For each fact × each AI, mark one of three: - Right — the AI got it right (or at least didn't contradict it) - Missing — the AI didn't mention it - Wrong — the AI said something that contradicts it
You'll end up with a grid. That grid is your entire to-do list for the next 90 days.
Prompt 3
Sort into Fix and Fill
Paste this into a fresh chat. Attach or paste your Facts list from Prompt 1 and the four AI answers from Prompt 2.
Here's my one-sentence claim, my facts list, and how four AIs answered six
questions about me today. Your job is to sort every gap into Fix or Fill,
using this rule:
Something's WRONG (an error) → FIX. The information exists but AI has it
wrong, missing, or contradicted. I need to correct every source that carries
it: my own site, my /llms.txt if I have one, and every external directory
(G2, Crunchbase, LinkedIn, Hugging Face, Wikipedia, press).
Something's MISSING (a hole) → FILL. There's no artifact for AI to cite. I
need to publish something only I could make. It must pass ALL FOUR tests:
- Only I could make it
- It compounds (a quarterly refresh keeps it fresh)
- It's quotable (a headline stat someone would cite)
- It points to my lane (not a generic explainer)
Fill types (pick one per Fill):
- A dataset — numbers only my product/customers generate
- A benchmark or index — I rank the market, I become the reference
- A simple tool — a calculator or checker buyers actually use
- Original research — findings worth citing
Return a table with these columns:
ID | Fix or Fill | What's wrong or missing | Sources to align (Fix) OR
Fill type + 4-test check (Fill) | Priority (P0/P1/P2) | Ship this week? Y/N
Rank Fix higher than Fill for the first quarter — Fix is what unlocks Fill's
citations. Cap at 8 Fixes for a first pass. Recommend exactly ONE Fill for
this quarter — the one that best passes all four tests.
What you get: your Fix list + your one Fill for the quarter. Paste into the Fix & Fill tab of the tracking sheet.
Prompt 4
Build one Fix ship kit
Do this once per Fix. Paste into a fresh chat.
I'm shipping FIX-[ID]: [one-line description].
My one-sentence claim: "[YOUR CLAIM FROM PROMPT 1]"
Here's what's wrong and where:
[paste the row from your Fix & Fill sheet, including sources to align]
Produce a Ship Kit I can execute in one sitting. Format as a single markdown doc:
1. Canonical sentence — the exact words every source will be aligned to.
Quoted, word-perfect.
2. Owned surfaces — for each of my own site pages that carries this fact:
- Page URL
- What the page says today (I'll paste this in if you can't fetch it)
- Exact replacement copy
- Any JSON-LD schema block to add (if applicable — write the full JSON)
3. My /llms.txt update — if I have an llms.txt, show me what to add or change.
If I don't have one, write me one from scratch, ~50-80 lines, covering:
canonical URL, one-sentence claim, product list, pricing summary,
canonical facts to cite, contact URL.
4. External directory drafts — for each external source (G2, Crunchbase,
LinkedIn company page, Hugging Face org, Wikipedia if I have an entry,
press outlets that covered me):
- Source name + URL
- How to submit the correction (support ticket / edit form / talk page /
correction email)
- Draft copy I can paste in, signed by me
5. Landed check — 3-5 things I can look for 7 days after shipping to
confirm the fix landed:
- "Grep my site for [key phrase] — should be present on N pages"
- "Check [external URL] — should say [expected string]"
- "Re-ask Prompt 2 question 3 in Claude — should not confuse me with
[confused brand]"
Keep these dead simple, no tools required.
6. Rollback — one sentence: what to do if the fix caused a problem.
Aim for one sitting to ship. If any owned surface needs more than a
copy-paste to update, flag it as "needs founder attention" and I'll handle
manually.
What you get: a Ship Kit per Fix. You do the fixes over your target ship weeks — one per week is a good pace. Every Ship Kit has a Landed check at day 7 so you can confirm the fix stuck.
Prompt 5
Brief your quarterly Fill
Do this once for your quarter's Fill. Paste into a fresh chat.
I'm building FILL: [name].
My one-sentence claim: "[YOUR CLAIM]"
Fill type: [dataset / benchmark or index / simple tool / original research]
The 4-test check first — answer YES to all four before we go further:
- Only I could make it? [YES/NO + why]
- It compounds? [YES/NO + why]
- It's quotable? [YES/NO + a candidate headline stat]
- It points to my lane? [YES/NO + how]
If any test is NO, tell me to pick a different Fill.
If all four are YES, produce a brief:
1. The one-page pitch — what this asset is, in ~150 words. Include the
candidate headline stat.
2. What data or content I need to make it — where does it come from, do I
already have it, or do I need to collect it? If I need to collect it,
what's the minimum viable version I could ship this quarter?
3. The report outline — 6-10 sections, 1,500-2,500 words total. Headline
stats up front, methodology at the end.
4. Charts I'll need — 4-8 chart ideas, each with title + what it shows +
what data feeds it.
5. Distribution assets — for each of these, give me a one-line description
of what it says:
- LinkedIn Article (800-1,200 words)
- Twitter/X thread (8-12 posts)
- Newsletter section (300-500 words)
- Homepage hero copy for launch week
- A short blurb for my /research or /blog index
6. Amplification queue — 13 weeks of angles on this same underlying asset.
One angle per week. Each angle is a fresh headline + a fresh chart or
data cut. If you can't come up with 13 without repeating, the Fill is
under-spec'd and I need a bigger data set.
7. Landed check — 30/60/90 day checks I can do without tools:
- Google the report title at day 7 — should appear in results
- Re-ask Prompt 2 questions at day 45 — at least 1/4 AIs should
reference the report
- Count backlinks manually at day 60 — should have at least 1 organic
What you get: the outline for your quarterly big asset + 13 weekly angles to amplify it. Now go write the report.
The tracking sheet
One source of truth, four tabs
Copy this Google Sheet template and use it as your single source of truth for the 90 days.
1. Facts
From Prompt 1. 30-50 rows. Columns: Section (Who I am / What I do / Who I'm for / Pricing / Unique / Where online) · Fact · Lives at (URL or 'nowhere') · On my site? Y/N · In external directory? Y/N
2. How AI answers you
From Prompt 2. One row per fact, one column per AI. Columns: Fact · ChatGPT (R/M/W) · Claude · Perplexity · Gemini · Wrong-count (formula) · Missing-count (formula)
3. Fix & Fill
From Prompt 3. Columns: ID · Type (Fix/Fill) · Description · Sources to align · Priority (P0/P1/P2) · Ship week · Status (Not started / Kit ready / Shipped / Landed) · Landed check date
4. Before & after
Rows: GSC clicks (last 90d) · GSC impressions (last 90d) · Right count total · Missing count total · Wrong count total · Named target ranking positions. Columns: Day 0 · Day 30 · Day 60 · Day 90 · Delta
The 90-day rhythm
What you do, week by week
-
·
Week 1
Claim + See
- –Prompts 1 and 2 done
- –Facts tab and How AI answers you tab filled in
- –Baseline (Day 0) numbers captured in the Before & after tab
-
·
Week 2
Sort
- –Prompt 3 done
- –Fix & Fill tab populated
- –6-8 Fixes scheduled across weeks 3-10; 1 Fill scheduled to ship in week 12
-
·
Weeks 3-10
Fix weekly
- –One Fix per week: run Prompt 4 → ship the Ship Kit → run the Landed check at day 7
- –Every Monday: re-run Prompt 2 question 3 (the confusion check) in all four AIs and mark any regression
-
·
Weeks 4-12
Fill quarterly
- –Week 4: run Prompt 5, get your Fill brief
- –Weeks 5-11: build the report, charts, and distribution assets
- –Week 12: ship the report + LinkedIn Article + Twitter thread + newsletter section on the same day
- –Week 13+: run the amplification queue, one angle per week
-
·
Day 30 / 60 / 90
Before & after
- –Re-run Prompt 2 in all four AIs; re-score
- –Pull fresh GSC clicks + impressions
- –Fill in the deltas
- –If Right count is up and Wrong/Missing is down, the lane is becoming yours
When to stop DIY
The DIY works. It also has a ceiling.
You should think about graduating to a done-for-you sprint (or hiring someone in-house to run this) when:
- →Your external directory drift check takes more than 30 minutes/week to run manually
- →You have >8 Fixes in flight and the manual re-run of Prompt 2 is slipping
- →Your Fill is the kind of quarterly asset that needs real data engineering — Cloudflare logs, GSC exports at scale, product telemetry
- →You want the ship kits generated, shipped, and verified without you assembling them
- →You want the Before & After numbers pulled automatically instead of tracked by hand
Design partners
Optimly is onboarding 20 design partners over the next 60 days. We'll run this playbook end-to-end for you, and measure outcomes.
We run this as a done-for-you 90-day sprint that includes this 360 audit, curated content assets, amplification across channels, and measurement that runs itself.
Book a discovery call →Credits
The framework — Claim your lane, Fix & fill the record, Find unclaimed questions, Answer on your site — is Optimly's. The one-sentence template, the four Fill tests, and the four Fill types are lifted directly from the Optimly playbook.
If this saved you time, share the playbook link with one other founder. That's how lanes get claimed.
