Every check, every point, every source.
This is Bellfound’s full method: every check, what it looks at, why it matters, how many points it is worth, and where the thinking comes from. The plain-words version is How we check.
How the score works.
- Points, not opinions. Every check has a fixed number of points. Your score is the points you earn out of the points that were checked.
- Only what ran counts. A check that couldn’t run, or is coming soon, never counts against you.
- The same page gets the same score. Where an AI reviewer is involved, it scores against a fixed checklist and the points come from fixed rules.
- Grades. Under 40% is Lost, 40–59% Wandering, 60–79% On the path, 80% and up Found.
- Not a ranking prediction. The score measures what is in your control. Nobody can honestly promise positions.
23 checks run today across seven areas; 4 more are coming soon.
01 Technical access · 15 points
Can Google and AI read your site?
Before anyone can find you, machines have to open your pages, be allowed in, and know which pages exist.
T1The page loads, with real text in it4 pts
- What we check
- We fetch your homepage the way a crawler does and check it answers, is a web page, and has readable text in the first response. Google can run JavaScript later; most AI crawlers don’t, so a page that only fills in with JavaScript gets a warning.
- Why it matters
- If machines can’t read your pages, nothing else on this report matters — you’re invisible at the front door.
- Points
- 4
- Source
- Google Search Central documentation
T2AI crawlers are allowed in3 pts
- What we check
- We read your robots.txt with the same longest-match rules engines use. Blocking the bots that fetch pages for live answers (for example OAI-SearchBot, ChatGPT-User, PerplexityBot, Claude-SearchBot, Bingbot) costs points. Blocking training-only bots (GPTBot, ClaudeBot, CCBot) is a fair choice: we note it and never take points for it.
- Why it matters
- Many sites accidentally block AI crawlers like ChatGPT’s — which silently removes you from AI answers entirely.
- Points
- 3
- Source
- OpenAI, Anthropic and Perplexity crawler documentation
T3A page list Google can follow3 pts
- What we check
- We find your sitemap, check it is valid, and read when it was last updated. Older than about six months is a warning.
- Why it matters
- Without a fresh sitemap, engines discover your pages late or not at all — new content can sit unseen for weeks.
- Points
- 3
- Source
- Google Search Central documentation
T4No page is hidden by mistake3 pts
- What we check
- We look for an accidental “noindex” in the page or its headers, which tells search engines to leave the page out.
- Why it matters
- One stray tag can remove an important page from Google completely, and nothing warns you.
- Points
- 3
- Source
- Google Search Central documentation
T5One clean address2 pts
- What we check
- We try the www and http versions of your address and expect each to arrive at one main address in a single step.
- Why it matters
- If both versions serve content separately, engines split your reputation between two “different” sites.
- Points
- 2
- Source
- Google Search Central documentation
02 Entity clarity · 12 points
Do they know who you are?
Machines need to match your website to the right business: your name, what you do, and where.
E1Who you are, in a form machines read5 pts
- What we check
- We look for structured data (Organization or LocalBusiness, plus WebSite) with your name, address of the site and logo. It helps engines tell who you are; it is not a ranking trick, and we don’t present it as one.
- Why it matters
- Without it, Google and AI engines have to guess who you are from scattered text. With it, you become a known entity — which feeds knowledge panels and helps AI name you correctly.
- Points
- 5
- Source
- Google Search Central documentation
E2Your name in your page titles4 pts
- What we check
- We take your business name from your site and check that your homepage and a sample of pages carry it in their titles.
- Why it matters
- When someone searches your brand, the title is the strongest signal connecting your name to your site. Missing it weakens every brand search.
- Points
- 4
- Source
- Google Search Central documentation
E3A real About page3 pts
- What we check
- We find your About page from your sitemap and links and measure how much real text it has. Engines answer “who is this business?” largely from here.
- Why it matters
- AI engines resolve your identity, history and credibility largely from this one page. A thin or missing one leaves you undefined.
- Points
- 3
- Source
- Google Search Central documentation
E4Your Google Business Profilecoming soon2 pts
- What we check
- Whether your business has a Google Business Profile that matches your site.
- Why it matters
- For local customers it is often the first thing engines show, before your website.
- Points
- 2 (not counted until it runs)
- Source
- Google Search Central documentation
E5The same name, address and phone everywherecoming soon1 pts
- What we check
- Whether your site and your Google listing agree on your name, address and phone number.
- Why it matters
- Inconsistent details make engines less confident you are one real business.
- Points
- 1 (not counted until it runs)
- Source
- Google Search Central documentation
03 Answer-ready structure · 15 points
Do your pages answer real questions?
Clear headings, steps and comparisons are the parts Google and AI lift straight into their answers.
S1Answers people can lift10 pts
- What we check
- An AI reviewer scores your pages against a fixed checklist: short definitions, steps, comparisons, lists or tables, and sections that stand on their own (2 points each). Points come from fixed rules, so the same page gets the same score. Having an FAQ block is not what counts.
- Why it matters
- AI engines lift self-contained sections as answers. Answer-shaped content gets quoted; prose-shaped content gets skimmed and skipped.
- Points
- 10
- Source
- Aggarwal et al., “GEO: Generative Engine Optimization”, KDD 2024
S2Clear headings5 pts
- What we check
- One clear main heading, no skipped levels, and headings that say what follows. More than one main heading is valid and loses a point only for being unclear, never called an error.
- Why it matters
- Honest sizing: Google tolerates messy headings, but AI engines rely on the outline to find liftable sections. One clear main heading (h1) per page, with a logical order, is basic readability for machines.
- Points
- 5
- Source
- Google Search Central documentation
04 Evidence · 20 points
Is there something worth quoting?
Answers quote pages that know what they are talking about: real numbers, sources and depth.
D1Real numbers, quotes and sources12 pts
- What we check
- An AI reviewer counts sourced numbers, named quotes and citations on your pages and flags keyword stuffing, against a fixed checklist. Points come from fixed rules on those counts, so the same page gets the same score. Research found pages with these were picked up by AI answers 30–40% more.
- Why it matters
- This is the biggest lever in AI search: engines cite content with sourced statistics and concrete facts 30–40% more. Marketing fluff is never quoted by anyone.
- Points
- 12
- Source
- Aggarwal et al., “GEO: Generative Engine Optimization”, KDD 2024
D2Enough depth, no copy-paste pages8 pts
- What we check
- We count the words on a sample of your pages and compare them to catch near-copies of the same template. It is about substance and duplicates, not a word-count target.
- Why it matters
- Thin or near-duplicate pages give engines nothing to rank and can actively hurt: they read as low-effort content.
- Points
- 8
- Source
- Google Search Central documentation
05 Off-site presence · 6 points
Does the rest of the web vouch for you?
Other places that mention and link to you tell machines you are a real business.
O2Your profiles, linked from your site2 pts
- What we check
- We look for links from your site to your social pages, Google Maps and directories, which machines follow to confirm you exist beyond your own website.
- Why it matters
- Those links let machines corroborate you against other sources; zero external profiles gives them nothing to check.
- Points
- 2
- Source
- Google Search Central documentation
O1Mentioned by independent sitesfull report or any plan4 pts
- What we check
- We search the web the way an AI assistant does and look for independent pages that mention you.
- Why it matters
- Machines trust what others say about you far more than what you say about yourself.
- Points
- 4 (counted in the full report and on every plan)
- Source
- Aggarwal et al., “GEO: Generative Engine Optimization”, KDD 2024
O3Sites that link to youcoming soon3 pts
- What we check
- A summary of the other websites that link to yours.
- Why it matters
- Engines still weigh who links to you heavily when deciding who deserves to be the answer.
- Points
- 3 (not counted until it runs)
- Source
- Google Search Central documentation
06 Enquiries · 15 points
Can a visitor reach you easily?
Being found only helps if people can contact you once they arrive.
C1Message you in one tap3 pts
- What we check
- A click-to-chat link (such as WhatsApp) a visitor can use straight away.
- Why it matters
- In Malaysia most enquiries arrive on WhatsApp. A number typed as plain text is a number nobody taps.
- Points
- 3
- Source
- Google Search Central documentation
C2Phone, address and hours2 pts
- What we check
- Your phone number, address and opening hours, readable on the page people contact you from.
- Why it matters
- A ready customer who has to retype your number usually does not, and an address stuck inside an image is invisible to AI.
- Points
- 2
- Source
- Google Search Central documentation
C3A way to enquire on every page3 pts
- What we check
- Whether the pages people land on from Google or AI answers each have a way to get in touch, not only the Contact page.
- Why it matters
- Every extra click between being convinced and getting in touch loses people who were ready.
- Points
- 3
- Source
- Google Search Central documentation
C4A price signal2 pts
- What we check
- A price, a range, a “from” price or an honest note on how pricing works. An AI assistant can’t quote a price that isn’t on the page.
- Why it matters
- Price silence is the most common reason a visitor leaves — and an AI assistant cannot quote a price you never published.
- Points
- 2
- Source
- Aggarwal et al., “GEO: Generative Engine Optimization”, KDD 2024
C5Proof you’re a real business2 pts
- What we check
- A company registration number or other people’s words about you, which visitors and AI engines both read.
- Why it matters
- A stranger about to message you is deciding whether to trust you, and AI engines quote reviews when they recommend.
- Points
- 2
- Source
- Google Search Central documentation
C6Speed of the page people enquire fromcoming soon3 pts
- What we check
- How fast that page feels on a phone. It needs real speed data, which we don’t collect yet.
- Why it matters
- A slow contact page loses the people furthest down the line, which are the most expensive visitors to lose.
- Points
- 3 (not counted until it runs)
- Source
- Google Search Central documentation
C7A short, easy form2 pts
- What we check
- Whether your contact form asks only for what it needs.
- Why it matters
- Every extra field costs you enquiries, and captchas turn away real people along with the bots.
- Points
- 2
- Source
- Google Search Central documentation
C8A promise to reply1 pts
- What we check
- A line such as “we reply within 2 hours”, so a visitor knows their message won’t vanish.
- Why it matters
- It reassures someone that their message is not vanishing into a void, and it sets the standard your team then keeps.
- Points
- 1
- Source
- Google Search Central documentation
07 AI visibility · 7 points
Does AI name you when customers ask?
We ask an AI assistant with web search what a customer would ask, and record whether your name is in the answer.
A1AI names you for what you dofull report or any plan4 pts
- What we check
- We ask the question a customer would ask about your kind of business and see whether you are named, and who is named instead.
- Why it matters
- If the answer names your competitor, they get the customer — and you never even know the question was asked.
- Points
- 4 (counted in the full report and on every plan)
- Source
- Aggarwal et al., “GEO: Generative Engine Optimization”, KDD 2024
A2AI knows who you arefull report or any plan3 pts
- What we check
- We ask about your business by name and check the answer describes you, not someone else with a similar name.
- Why it matters
- An AI that cannot establish who you are will never recommend you — unknown means invisible.
- Points
- 3 (counted in the full report and on every plan)
- Source
- Aggarwal et al., “GEO: Generative Engine Optimization”, KDD 2024
Sources.
- Google Search Central: JavaScript, robots.txt, sitemaps, titles, structured data, AI featuresdevelopers.google.com/search
- Aggarwal et al., “GEO: Generative Engine Optimization”, KDD 2024arxiv.org/abs/2311.09735
- OpenAI crawler documentation (GPTBot, OAI-SearchBot, ChatGPT-User)platform.openai.com/docs/bots
- Anthropic crawler documentation (ClaudeBot, Claude-User, Claude-SearchBot)support.anthropic.com
- Perplexity crawler documentation (PerplexityBot, Perplexity-User)docs.perplexity.ai