Twelve checks, in four layers, that decide whether an answer engine can assemble your brand into a short list. Work through it on screen or print it. Nothing is sent anywhere; the boxes live in this browser tab only.
Can an engine get your words at all? Everything above this layer is wasted if it cannot.
Grep your server or CDN logs for GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot and ClaudeBot. Exclude the homepage.
Pass: at least one fetch of a deeper page. Homepage-only crawling is the most common failure and it looks like success on a dashboard.
Run curl -s https://yoursite.com/page | grep "your sentence". No browser, no rendering.
Pass: the sentence comes back. If it does not, the crawler that skips rendering never sees your argument.
Read your own robots.txt line by line. Write down, for each AI agent, whether the block was a decision or an inheritance from a template.
Pass: every rule is deliberate. A blanket block copied from a boilerplate file is a fail even if traffic looks fine. Google-Extended belongs in this check and not in R1: it is a robots.txt token governing Gemini grounding, not a user agent that shows up in your logs.
Click from the homepage to the page you most want cited. Count the clicks and the URL folder depth.
Pass: two clicks or fewer. Pages four folders deep draw close to zero AI referral traffic.
Does the engine connect you to the category, or only to your own name? Buyers search the category.
Ask five engines the query your buyer would type, with your company name removed. Record whether you are named. Repeat on a different day.
Pass: named in at least one of five, on both runs.
Ask each engine what your company does. Read only the first sentence.
Pass: the first sentence names your actual category and names no wrong one. A hedge or a stale category is a fail.
List your page titles. Look for the category noun your buyer uses, not your internal product name.
Pass: at least one page, other than the homepage, carries the buyer's term in its title and its first paragraph.
A claim that exists only on your own domain has nothing holding it up.
Search the exact phrasing of the claim. Count distinct domains, excluding yours and any you control.
Pass: three or more, each reachable and each dated.
Check whether one of the three is a reference site, a discussion forum, a video platform, a review platform, or a trade publication.
Pass: yes. Wikipedia, YouTube and Reddit alone account for 15% of the sources listed in Google's AI summaries.
Find listicles and comparison pages for your category. Check the publication date and whether you are in the list.
Pass: at least one, published within 12 months, not written or paid for by you.
The answer is two to four names long. Your passage has to survive being lifted out of its page.
Copy the passage you want quoted into an empty document. Read it cold, with no heading and no preceding paragraph.
Pass: it still answers the question. If it needs the sentence before it, it cannot be lifted.
Underline every claim in the passage. Strike out anything an adjective is doing the work for.
Pass: at least one number, constraint, date or price survives, and it names both you and the category.
| Score | Band | What it means |
| 0 to 3 | Invisible | The engine cannot assemble you. Fix Layer 1 before anything else; work above a broken retrieval layer does not compound. |
| 4 to 6 | Retrievable, not selectable | Your words are reachable but nothing attaches them to the category. Layer 2 is the constraint. |
| 7 to 9 | In the running | You get named some of the time. Corroboration and passage shape decide the rest. |
| 10 to 12 | Citable | Now measure. An audit tells you that you can be assembled, never how often you actually are. |