▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲▲
║ . . : : ░ ▒ ▓ ▓ SCANNING… ▓ ▓ ▒ ░ : : . . ║
║ : : ░ ▒ ▓ ▓ ░▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒░ ▓ ▓ ▒ ░ : : ║
║ →→→ █ █ ░ ░ FULL ER-SCAN OVERVIEW ░ ░ █ █ ←←← ║
║ : : ░ ▒ ▓ ▓ ░▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒░ ▓ ▓ ▒ ░ : : ║
▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼▼
POKPOK.AI // PRE-LAUNCH CHEAT SHEET
AI Visibility Checklist
The 5 AI metrics and all 22 sub-metrics — for founders who don’t have a website yet.
This is the companion to the HI (Human Impression) checklist. Where HI measures how a human reads your site, the AI metrics measure how AI answer engines — ChatGPT, Perplexity, Claude, Google AI Overviews — discover, read, understand, verify and act on your site.
Getting this right at the planning stage costs almost nothing. Retrofitting it later means re-architecting pages. Every sub-metric below lists what it measures, why it matters, and what to decide before you build anything.
05
AI METRICS
22
SUB-METRICS
65
DECISIONS TO MAKE
€0
COST TO PLAN IT IN
AI01 // FIND
AI Findability
//AI01
AI FIND · Findability
[!] Core question
Can AI systems discover and fetch your site at all?
[?] WHY IT MATTERS
This is the base layer. If crawlers can't reach or fetch your pages, nothing else in this framework matters — you're invisible to every AI answer engine before any content quality is even evaluated.
FINDABILITY SUB-METRICS
12 DECISIONS
AI01-01
3 DECISIONS
Crawler Access
WHAT IT MEASURES
Whether your robots.txt explicitly welcomes AI crawlers (GPTBot, PerplexityBot, ClaudeBot, Google-Extended, OAI-SearchBot, CCBot) or accidentally blocks them.
[?] WHY IT MATTERS
Many site builders and “AI-blocking” plugins silently disallow these bots by default. You can lose 100% of AI visibility without ever noticing.
BEFORE LAUNCH, DECIDE
[ ]
Publish a robots.txt that explicitly allows the major AI crawlers — unless you have a deliberate reason to block training or retrieval.
[ ]
Know the difference between blocking training crawlers and blocking retrieval/answer crawlers. You may want one and not the other.
[ ]
Re-check after every platform migration. New templates often reset robots rules.
AI01-02
3 DECISIONS
Machine Map Access
WHAT IT MEASURES
Whether you have a clean, valid XML sitemap, and whether the routes it lists actually resolve without errors.
[?] WHY IT MATTERS
Crawlers use your sitemap to decide which pages exist and matter. A messy or incomplete map means large parts of your site are invisible or under-prioritised.
BEFORE LAUNCH, DECIDE
[ ]
Generate and submit a valid sitemap.xml from day one, even with just 5–10 pages.
[ ]
Check periodically that sitemap URLs return clean 200 codes — no dead links, no redirect chains.
[ ]
Consider an optional llms.txt: a short guide telling AI systems which pages matter most. Cheap to add, rarely done, so it is a quick differentiator.
AI01-03
3 DECISIONS
Canonical Health
WHAT IT MEASURES
Whether every page has one clear, authoritative URL (canonical tag) with no duplicate or conflicting versions splitting your authority.
[?] WHY IT MATTERS
If www.site.com, site.com and site.com/index all serve slightly different signals, AI systems cannot tell which is authoritative — diluting your visibility.
BEFORE LAUNCH, DECIDE
[ ]
Pick ONE canonical domain format (with or without www) and enforce HTTPS everywhere.
[ ]
Set canonical tags correctly on every page template before content scales.
[ ]
Avoid cross-domain canonicals unless intentional, e.g. syndicated content.
AI01-04
3 DECISIONS
Fetchability
WHAT IT MEASURES
Whether the page actually loads clean content when requested — no CAPTCHAs, login walls, interstitials or bot-blocking services standing in the way.
[?] WHY IT MATTERS
Even with perfect robots.txt rules, aggressive bot-protection services (Cloudflare, DataDome) can still silently block legitimate AI crawlers.
BEFORE LAUNCH, DECIDE
[ ]
If you use a CDN or bot-protection service, explicitly allow-list known AI crawler user-agents.
[ ]
Avoid interstitials, cookie walls and age-gates on your core marketing pages.
[ ]
Test your live site with a plain fetch, no JavaScript. If what comes back is thin or blank, you have a rendering dependency crawlers may not solve.
AI02 // READ
AI Readability
//AI02
AI READ · Readability
[!] Core question
Once an AI system fetches your page, can it correctly parse and understand the structure?
[?] WHY IT MATTERS
Fetching isn't reading. A page can be 100% accessible but so poorly structured that a machine extracts garbage or misses your core message entirely.
READABILITY SUB-METRICS
12 DECISIONS
AI02-01
3 DECISIONS
Text Extraction
WHAT IT MEASURES
What share of your page’s visible content survives clean text extraction, versus being trapped in images, JavaScript-only rendering or cookie banners.
[?] WHY IT MATTERS
Whatever an AI can extract is the entirety of what it knows about you. If your value proposition lives only in an image or a JS widget, it effectively does not exist.
BEFORE LAUNCH, DECIDE
[ ]
Keep core marketing copy as real, server-rendered text — not baked into images, not requiring JavaScript to appear.
[ ]
Minimise cookie-consent banner text competing with real content in the extracted text.
[ ]
If you use React, Vue or similar, confirm server-side rendering or static generation is enabled for public pages.
AI02-02
3 DECISIONS
Structure Clarity
WHAT IT MEASURES
Whether your heading hierarchy is complete, logically ordered, and gives a machine a clean outline of the page.
[?] WHY IT MATTERS
AI systems use heading structure the way a human uses a table of contents. A broken hierarchy makes your page harder to summarise accurately.
BEFORE LAUNCH, DECIDE
[ ]
Plan exactly ONE H1 per page — your core headline.
[ ]
Structure supporting sections as properly nested H2s and H3s. Never skip levels, never use headings purely for visual styling.
[ ]
Draft your content outline as a heading tree before writing prose. It forces logical structure early.
AI02-03
3 DECISIONS
Content Frontload
WHAT IT MEASURES
How many words into the page your brand name and core offer first appear.
[?] WHY IT MATTERS
AI summarisation leads with whatever appears first. If your brand name appears 100+ words in, AI-generated summaries about you lead with generic category language instead of your name.
BEFORE LAUNCH, DECIDE
[ ]
Name your brand AND state your core offer within the first sentence or two of your homepage. Resist the broad, content-led hook.
[ ]
Write your meta description to front-load brand and offer too — it is often quoted directly.
[ ]
Avoid content-marketing openings on money pages. Save the broad industry statement for blog posts.
AI02-04
3 DECISIONS
Hidden Content Risk
WHAT IT MEASURES
Whether meaningful content is trapped inside carousels, accordions, modals or other interactive elements that may not render in a simple crawl.
[?] WHY IT MATTERS
Content that needs a click or a slide to appear may be invisible to AI crawlers, even when it is perfectly visible to humans.
BEFORE LAUNCH, DECIDE
[ ]
Never put unique, non-repeated information exclusively inside a carousel or accordion.
[ ]
If a carousel is needed for design reasons, make sure the same copy also exists statically in the page source.
[ ]
Add descriptive alt text to every meaningful image. It is free content-extraction insurance.
AI03 // KNOW
Knowledge
Does AI understand who you are, what you offer, and where you fit in the marketplace?
THE TRUTH
You cannot be recommended until you are understood.
THE FIX
Say plainly who you are — in the copy and in the schema.
KNOWLEDGE SUB-METRICS
14 DECISIONS
AI03-01
3 DECISIONS
Entity Clarity
WHAT IT MEASURES
Whether structured data (Organization schema / JSON-LD), page titles, Open Graph tags and visible branding all agree on your exact brand name with no conflicts.
[?] WHY IT MATTERS
If your JSON-LD says “Acme Inc.”, your title tag says “Acme” and your OG tags say something else, AI has to guess which is authoritative — reducing confidence in every fact tied to your brand.
BEFORE LAUNCH, DECIDE
[ ] Implement basic Organization schema (JSON-LD) from day one. Small dev task, outsized clarity benefit.
[ ] Spell your brand name identically across title tag, schema, Open Graph, footer and logo alt text.
[ ] Link official social profiles via sameAs in your schema. It helps AI cross-verify you are a real entity.
AI03-02
2 DECISIONS
Offer Clarity
WHAT IT MEASURES
Whether what you offer, who it is for, and what outcome you deliver are all stated in plain, unambiguous language somewhere on the page.
[?] WHY IT MATTERS
AI answering “what does this company do” pulls directly from explicit statements. Vague or metaphor-heavy copy makes you much harder to summarise correctly.
BEFORE LAUNCH, DECIDE
[ ] Write one plain sentence: what you offer + who it is for + what outcome they get. Put it near the top, even if the hero headline is more creative.
[ ] Never rely on a clever abstract tagline alone to carry your offer. Pair it with a literal one-liner.
AI03-03
3 DECISIONS
Category Clarity
WHAT IT MEASURES
Whether your industry and category signals are consistent across copy, schema and page structure, so a machine can shelve you next to the right competitors.
[?] WHY IT MATTERS
Miscategorisation means you get excluded from comparison answers in your actual category, or wrongly included in irrelevant ones.
BEFORE LAUNCH, DECIDE
[ ] State your category in plain language — “B2B SaaS brand perception platform” — rather than relying on buzzwords.
[ ] Make category signals in your schema markup match the category language in your visible copy.
[ ] Check how competitors describe their category, so you do not use terminology that files you somewhere else.
AI03-04
3 DECISIONS
Brand Name Anchoring
WHAT IT MEASURES
Whether your brand exists in public knowledge graphs (Wikidata, Google Knowledge Graph), and whether your name is unique enough to avoid collision with other entities.
[?] WHY IT MATTERS
A brand with zero external anchoring is essentially unverified. AI has no prior record you exist beyond your own website’s claims.
BEFORE LAUNCH, DECIDE
[ ] Choose a genuinely distinctive brand name. Check for collisions with existing companies before you commit.
[ ] Do not chase a Wikipedia page early — it is earned, not requested. Build real citations across independent platforms instead.
[ ] Register consistent business listings (Crunchbase, LinkedIn Company Page, G2 or Capterra if relevant) early. They feed knowledge-graph resolution over time.
AI03-05
3 DECISIONS
Quotable Facts
WHAT IT MEASURES
Whether your page contains durable, neutral, factual statements about the business — not just performance claims — that AI can quote long after specific stats age out.
[?] WHY IT MATTERS
“Grew traffic 20%” is stale within a year. “Company X is a [category] founded in [year], serving [audience]” stays quotable indefinitely.
BEFORE LAUNCH, DECIDE
[ ] Write at least one plain, permanent, non-marketing sentence defining what your company is.
[ ] Separate durable facts (founding year, category, audience) from time-bound performance claims, and plan to refresh the latter.
[ ] Use named entities — specific numbers, dates, named clients — instead of vague superlatives. Specificity is what makes a sentence quotable.
AI03 // KNOW
Knowledge
//AI01
AI FIND · Findability
[!] Core question
Can AI systems discover and fetch your site at all?
[?] WHY IT MATTERS
This is the base layer. If crawlers can't reach or fetch your pages, nothing else in this framework matters — you're invisible to every AI answer engine before any content quality is even evaluated.
KNOWLEDGE SUB-METRICS
14 DECISIONS
AI01-01
3 DECISIONS
Crawler Access
WHAT IT MEASURES
Whether your robots.txt explicitly welcomes AI crawlers (GPTBot, PerplexityBot, ClaudeBot, Google-Extended, OAI-SearchBot, CCBot) or accidentally blocks them.
[?] WHY IT MATTERS
Many site builders and “AI-blocking” plugins silently disallow these bots by default. You can lose 100% of AI visibility without ever noticing.
BEFORE LAUNCH, DECIDE
[ ]
Publish a robots.txt that explicitly allows the major AI crawlers — unless you have a deliberate reason to block training or retrieval.
[ ]
Know the difference between blocking training crawlers and blocking retrieval/answer crawlers. You may want one and not the other.
[ ]
Re-check after every platform migration. New templates often reset robots rules.
AI01-02
3 DECISIONS
Machine Map Access
WHAT IT MEASURES
Whether you have a clean, valid XML sitemap, and whether the routes it lists actually resolve without errors.
[?] WHY IT MATTERS
Crawlers use your sitemap to decide which pages exist and matter. A messy or incomplete map means large parts of your site are invisible or under-prioritised.
BEFORE LAUNCH, DECIDE
[ ]
Generate and submit a valid sitemap.xml from day one, even with just 5–10 pages.
[ ]
Check periodically that sitemap URLs return clean 200 codes — no dead links, no redirect chains.
[ ]
Consider an optional llms.txt: a short guide telling AI systems which pages matter most. Cheap to add, rarely done, so it is a quick differentiator.
AI01-03
3 DECISIONS
Canonical Health
WHAT IT MEASURES
Whether every page has one clear, authoritative URL (canonical tag) with no duplicate or conflicting versions splitting your authority.
[?] WHY IT MATTERS
If www.site.com, site.com and site.com/index all serve slightly different signals, AI systems cannot tell which is authoritative — diluting your visibility.
BEFORE LAUNCH, DECIDE
[ ]
Pick ONE canonical domain format (with or without www) and enforce HTTPS everywhere.
[ ]
Set canonical tags correctly on every page template before content scales.
[ ]
Avoid cross-domain canonicals unless intentional, e.g. syndicated content.
AI01-04
3 DECISIONS
Fetchability
WHAT IT MEASURES
Whether the page actually loads clean content when requested — no CAPTCHAs, login walls, interstitials or bot-blocking services standing in the way.
[?] WHY IT MATTERS
Even with perfect robots.txt rules, aggressive bot-protection services (Cloudflare, DataDome) can still silently block legitimate AI crawlers.
BEFORE LAUNCH, DECIDE
[ ]
If you use a CDN or bot-protection service, explicitly allow-list known AI crawler user-agents.
[ ]
Avoid interstitials, cookie walls and age-gates on your core marketing pages.
[ ]
Test your live site with a plain fetch, no JavaScript. If what comes back is thin or blank, you have a rendering dependency crawlers may not solve.
AI01-04
3 DECISIONS
Fetchability
WHAT IT MEASURES
Whether the page actually loads clean content when requested — no CAPTCHAs, login walls, interstitials or bot-blocking services standing in the way.
[?] WHY IT MATTERS
Even with perfect robots.txt rules, aggressive bot-protection services (Cloudflare, DataDome) can still silently block legitimate AI crawlers.
BEFORE LAUNCH, DECIDE
[ ]
If you use a CDN or bot-protection service, explicitly allow-list known AI crawler user-agents.
[ ]
Avoid interstitials, cookie walls and age-gates on your core marketing pages.
[ ]
Test your live site with a plain fetch, no JavaScript. If what comes back is thin or blank, you have a rendering dependency crawlers may not solve.
AI04 // TRUST
Trust
Can AI verify what you claim, and does it have reason to believe you are a current, legitimate, accountable source?
THE TRUTH
This is usually the hardest metric to score well on early, and the one that separates technically well-built sites from sites AI systems actually trust enough to cite or recommend.
THE FIX
Attach a number or a date to every claim, link every claim to proof, date every page, collect reviews on two or three independent platforms.
TRUST SUB-METRICS
12 DECISIONS
AI04-01
3 DECISIONS
Fact Density
WHAT IT MEASURES
How many verifiable, specific data points — numbers, dates, named entities — appear per 500 words of copy, versus vague marketing language.
[?] WHY IT MATTERS
Machines repeat and cite specific, checkable facts far more readily than generic claims. Low fact density means you get summarised in generic terms instead of your own words.
BEFORE LAUNCH, DECIDE
[ ] For every benefit claim, ask: can I attach a real number or date to this? Replace vague language wherever you can.
[ ] Include named entities — real client names with permission, specific dates, specific metrics — not “many clients” or “significant growth”.
[ ] Avoid copy that is 100% adjectives with zero nouns or numbers behind them.
AI04-02
3 DECISIONS
Claim Checkability
WHAT IT MEASURES
Whether your claims are technically verifiable, and whether they carry direct proof links — case studies, source pages — an AI can follow to confirm them.
[?] WHY IT MATTERS
A claim that is technically true but has no supporting link is treated with more caution than one linked to evidence. Machines, and sceptical humans, hedge on unlinked claims.
BEFORE LAUNCH, DECIDE
[ ] Every performance claim links directly to its source: a case study, a review, a dated report. Plan that link structure into your templates from the start.
[ ] Do not make claims you have no intention of ever backing with a case study or evidence page.
[ ] Certifications link to the verifying body, not just a badge image.
AI04-03
3 DECISIONS
Source Markers
WHAT IT MEASURES
Whether pages carry publication dates, last-modified dates, and author or attribution information.
[?] WHY IT MATTERS
Without dates or authorship a machine cannot tell whether a page is current or stale, or who stands behind the claims. One of the cheapest fixes with outsized trust impact.
BEFORE LAUNCH, DECIDE
[ ] Build published date and last-updated date into every page template from day one — even a simple footer line. Retrofitting across dozens of pages is tedious.
[ ] Add author attribution to blog and educational content, even if it is only “Written by the [Company] team”.
[ ] Keep legal, contact and privacy pages present and easy to find. Their presence is itself a baseline trust signal.
AI04-04
3 DECISIONS
External Validation
WHAT IT MEASURES
Whether your brand is corroborated by independent third parties — review platforms, press mentions, community discussion — beyond your own website’s claims.
[?] WHY IT MATTERS
For most B2B categories this is the single largest trust gap. AI weights independent corroboration heavily; a brand that only vouches for itself ranks below one confirmed by outside sources.
BEFORE LAUNCH, DECIDE
[ ] Pick 2–3 independent review platforms relevant to your category and start collecting reviews with your first handful of customers. Do not rely on one platform.
[ ] Plan a lightweight PR or community presence — guest posts, podcasts, forums. Any independent mention helps.
[ ] Do not wait for volume. Even 5–10 genuine reviews on one platform beats zero by a wide margin.
AI05 // DO
Action
Once AI, or a human, understands you, can it guide someone all the way to taking action?
THE TRUTH
This is where interest converts to a lead or a sale. A site can be perfectly findable, readable and trustworthy and still fail here if the path to action is unclear or broken.
THE FIX
Pick one call to action, test every link on its path, describe the offer in structured data, remove login walls and modal traps.
ACTION SUB-METRICS
15 DECISIONS
AI05-01
3 DECISIONS
Action Intent
WHAT IT MEASURES
Whether there is ONE consistent, clearly signalled primary action used uniformly across the whole page — not several competing calls to action.
[?] WHY IT MATTERS
AI systems, and humans, need one unambiguous answer to “what happens if I click”. Multiple inconsistent CTAs dilute that signal.
BEFORE LAUNCH, DECIDE
[ ] Choose ONE primary action for the entire site — “Book a Call”, “Start Free Trial” — before designing any page.
[ ] Use identical CTA wording everywhere. Do not let “Book a Call”, “Let’s Talk” and “Get Started” coexist as if equally primary.
[ ] Use clear action verbs — book, start, get, try — not vague ones like explore, learn, discover.
AI05-02
3 DECISIONS
Action Route
WHAT IT MEASURES
Whether every path connected to your primary action — contact forms, booking links, nav links, footer links — resolves cleanly with no redirects, dead ends or broken links.
[?] WHY IT MATTERS
One broken link on the conversion path silently kills a meaningful share of leads. AI agents trying to complete the action on a user’s behalf simply fail and move to a competitor.
BEFORE LAUNCH, DECIDE
[ ] Test every link on the CTA path — nav, footer, contact, booking — before launch and after every site update.
[ ] Avoid redirect chains. Link directly to final destinations.
[ ] If you use a booking tool, confirm the embed works outside your own browser session. Test it in incognito.
AI05-03
3 DECISIONS
Offer Data
WHAT IT MEASURES
Whether structured data (Schema.org Service or Offer markup) describes your service tiers, eligibility, pricing and availability in a machine-readable format.
[?] WHY IT MATTERS
Without structured offer data, AI agents cannot summarise or compare your terms. They can only say “go book a call”, skipping the pre-qualification or comparison step that could favour you.
BEFORE LAUNCH, DECIDE
[ ] Plan to add basic Service or Offer schema once you have defined service tiers. Even simple markup beats none.
[ ] Decide your pricing disclosure approach — full price, range or eligibility criteria — and encode it as structured data, not just prose.
[ ] If eligibility requirements exist, state them explicitly rather than leaving buyers to self-select.
AI05-04
2 DECISIONS
Task Fit
WHAT IT MEASURES
Whether your primary action matches what visitors in your category actually expect — a B2B service site is expected to offer contact or booking, not “buy now”.
[?] WHY IT MATTERS
An action that mismatches category expectations confuses AI systems and human visitors alike about what kind of business you are.
BEFORE LAUNCH, DECIDE
[ ] Research the standard primary action for your category and business model before choosing yours.
[ ] Make CTA language match the actual next step. Do not say “Get a Quote” if it opens a generic contact form with no quoting mechanism.
AI05-05
4 DECISIONS
Action Blockers
WHAT IT MEASURES
Whether anything technical stands between intent and completion: broken links, login walls, JavaScript-only forms, cookie walls, soft-404s or modal-only actions.
[?] WHY IT MATTERS
One hard blocker on your only conversion path can cap conversion at zero for a whole segment of visitors, and completely block AI agents attempting the action programmatically.
BEFORE LAUNCH, DECIDE
[ ] No login walls on any page or form meant to generate leads. Save gating for post-conversion content.
[ ] Test your primary form with JavaScript disabled. Pure JS-only forms are invisible to some crawlers and slower for many users.
[ ] Never trap your action exclusively inside a modal with no static fallback link.
[ ] Run a basic broken-link check before launch and periodically afterwards.
06 // QUICK REFERENCE
All 22 sub-metrics at a glance
AI Find
//AI01-01
Crawler Access
Allow AI bots in robots.txt
//AI01-02
Machine Map Access
Valid, complete sitemap
//AI01-03
Canonical Health
One authoritative URL per page
//AI01-04
Fetchability
No CAPTCHAs/walls blocking bots
AI Read
//AI02-01
Text Extraction
Real text, not images/JS-only
//AI02-02
Structure Clarity
Clean H1→H2→H3 hierarchy
//AI02-03
Content Frontload
Brand + offer in first sentence
//AI02-04
Hidden Content Risk
Nothing critical trapped in carousels
AI Know
//AI03-01
Entity Clarity
Allow AI bots in robots.txt
//AI03-02
Offer Clarity
Valid, complete sitemap
//AI03-03
Category Clarity
One authoritative URL per page
//AI03-04
Brand Name Anchoring
No CAPTCHAs/walls blocking bots
//AI03-04
Quotable Facts
No CAPTCHAs/walls blocking bots
AI Trust
//AI04-01
Fact Density
Allow AI bots in robots.txt
//AI04-02
Claim Checkability
Valid, complete sitemap
//AI04-03
Source Markers
One authoritative URL per page
//AI04-04
External Validation
No CAPTCHAs/walls blocking bots
AI Do
//AI05-01
Action Intent
Allow AI bots in robots.txt
//AI05-02
Action Route
Valid, complete sitemap
//AI05-03
Offer Data
One authoritative URL per page
//AI05-04
Task Fit
No CAPTCHAs/walls blocking bots
//AI05-04
Action Blockers