Bramblewick Plumbing & Drain does not exist. It is a business we invented so that we could publish a complete audit without publishing a client's private document. The findings are real findings of the kind we produce; the site, the numbers, the reviews and the person named on this cover are not. Nothing here describes a real company, and no client's report has ever been published.
Everything else about the document is exactly what a client receives — the same structure, the same evidence blocks, the same refusals.
A published rule from Google or Bing. We can show you the sentence. Breaking it risks something concrete — a suspended profile, a page removed.
Our professional judgement, not a published rule. Where a policy says what is forbidden but not what to do instead, the "what to do instead" is ours and is labelled as ours.
Found and recorded, but we are not acting on it yet, and the reason is stated on the item.
Every finding carries a line like this in our file, and it is reproduced here rather than hidden. In order: which lens found it · what it rests on, with the pointer to the rule · what was measured, and on what date.
DOC means published documentation. GATE means it also trips our pre-ship checklist. A finding that rests only on professional opinion never becomes something we ask you to do — you will not find one on your list.
No score, no grade, no "73 out of 100", no letter, no traffic-light dial. Those numbers are invented by whoever sells them and there is no published basis for any of them. We would rather give you fifteen things you can act on.
No count of "212 errors". A crawler will happily produce one; almost none of it matters, and burying the fifteen things that do inside two hundred that don't is not a service.
Two things, both on item 15 and item 9: photographs of three recent jobs with the name of the technician who did them, and about ten minutes on a call about your Business Profile categories. Nothing else on this list waits on you.
Everything we say about your business — your services, your licences, your hours — comes from what you told us, never from what we guessed.
| Where you might be found | Controls being seen | Controls training | Your site today | |
|---|---|---|---|---|
| 1 | Google AI Overviews & AI Mode | Search Console → Search generative AI | Google-Extended | Blocked Your property is set to Inherit; the parent property above it is set to Exclude. Nobody chose this for your site. |
| 2 | Google Gemini apps | — no separate control published | Google-Extended | Grounding off Google-Extended is disallowed in your robots.txt. Google states this does not affect Google Search. It does switch off grounding in Gemini apps. |
| 3 | Microsoft Copilot (Bing) | Bingbot access; no noarchive | — none published | Degraded All 46 pages carry noarchive from your theme. Google ignores that tag. Bing documents it as preventing your content being used in Copilot. |
| 4 | ChatGPT search | OAI-SearchBot | GPTBot | Blocked OAI-SearchBot is disallowed. OpenAI: opted-out sites “will not be shown in ChatGPT search answers”. |
| 5 | Perplexity | PerplexityBot | — not used for training | Blocked PerplexityBot is disallowed. Perplexity recommends allowing it and says it is not used to train models. |
| 6 | Claude | Claude-SearchBot and Claude-User | ClaudeBot | Clear No rule in your file names any of Anthropic’s three robots. Claude can see your site. |
| 7 | Meta AI | Meta-WebIndexer | Meta-ExternalAgent | Clear No rule names Meta-WebIndexer. Meta: allowing it “helps us cite and link to your content in Meta AI’s responses”. |
| 8 | Apple Siri, Spotlight, Safari | Applebot, plus nosnippet for AI answers | Applebot-Extended | Clear today No rule names Applebot, and you carry no Googlebot restriction — which matters, because Apple follows Googlebot’s rules when its own are absent. See overleaf. |
Three blocked, two degraded, three clear — and the three that are clear are clear by luck. Your robots.txt names five robots. Nothing in it mentions Anthropic’s, Meta’s or Apple’s, so those three platforms can see you — not because anyone decided they should, but because the list somebody pasted in did not happen to include them. That is worth knowing before the list gets “updated” by someone trying to help.
Set the Search Console control to Include and check the parent property. Remove the two disallow lines that block ChatGPT and Perplexity from seeing you. Remove the noarchive tag, which costs you on Bing and does nothing on Google. Leave Claude, Meta and Apple alone — and record that they are deliberately allowed, so the next person to touch the file knows.
Five lines of work. Most of it is one afternoon.
Three robots in your file — GPTBot, Google-Extended and CCBot — govern whether your content may be used to train AI models. That is a business decision about your own words and photographs, and it is yours. We have not touched them and will not without being asked.
If you want them left exactly as they are, that is a complete answer and it costs you nothing documented.
Your theme adds a noarchive tag to every page. Its close relative, nosnippet, is the single highest-consequence line in this whole area, because three different companies document three different effects for it: Google treats it as removing your page as a direct input to AI Overviews and AI Mode; Bing says it may limit Copilot citation quality; and Apple documents it as the opt-out from its AI answers — a page carrying it is not used as context when Apple generates an answer in Siri or Search.
You do not carry nosnippet today. You carry noarchive, which is the one Google stopped using and Bing still acts on. We are removing it, and we are checking every page for nosnippet at the same time, because a theme or plugin update can add one without anyone noticing.
What we will record for you, and why it matters more than the fix. After we change your file, we will leave a short comment in it naming each platform, which robot governs what, and which choices were yours. The next person who touches that file — a developer, a plugin, a well-meaning forum post — will see it. The fix takes an afternoon; the note is what stops it being undone in a year.
| What we will report | Where it comes from | What it does and does not tell you |
|---|---|---|
| AI impressions, by page | Search Console → Generative AI performance report · monthly | How often your pages were linked from Google’s AI Overviews and AI Mode, by page, country, device and date. It does not show clicks, click-through rate, position, or which questions triggered it — Google does not publish those. |
| Arrivals from AI assistants | Google Analytics → “AI Assistant” channel · monthly | People who reached your site from ChatGPT, Gemini, Copilot and similar. Google’s definition explicitly excludes its own AI Overviews and AI Mode, whose clicks sit in your ordinary search totals. Arrival, not citation. |
| Citation checks | A person, in a browser · monthly | Someone types your priority questions into each assistant and records whether you were named, by which page, and who was named instead. Date, device, location and signed-out state go on every row. |
| Nothing else | — | We do not use software to query Google or any AI assistant on your behalf. Google’s rules prohibit automated querying without permission, and we hold the assistants to the same standard. |
A citation you see once is not a thing you have. Google states that AI Overviews “are only shown when our systems determine that it is additive to classic Search, and as such, often don’t trigger”, and that its two AI features “may use different models and techniques” so the links they show will vary. Every check we run is one reading under one set of conditions, and it is labelled as one.
All eight companies publish which of their robots can reach your site and what blocking each one does. So we can tell you definitively whether each AI system is able to see you.
None of them publishes how its AI answers choose which sources to cite. We looked, on every page, and recorded where we looked. So nobody — us, or anyone selling you “AI SEO” — can tell you that you will appear. Anyone who says otherwise is guessing with your money.
One partial exception, for completeness: Apple publishes five factors it says it may take into account when ranking its own web search results, explicitly with no stated order of importance. That is Apple Search, not AI answers, and we treat it as nothing more than it says.
“For Google Search, you can ignore tactics like “chunking” content, creating unnecessary AI text files (like llms.txt), or pursuing inauthentic mentions.”
Google — AI optimization guide, doc updated 2026-07-10 · read 2026-09-11No llms.txt, no “AI schema”, no chopping your pages into chunks, no writing in a special voice for machines, no bought mentions, no page per variation of a question. Meta adds its own: it prefers “industry-standard practices like robots.txt rather than non-standard formats like NoAI tags.” And we will never hide instructions to an AI inside your pages — Bing documents that as prohibited, and we treat it as a line we do not go near.
“Just because a page meets all requirements, best practices, and complies with the policies, doesn’t mean that Google will crawl, index, or serve its content. Indexing and serving aren’t guaranteed.”
Google — AI optimization guide · read 2026-09-11We are quoting that against our own interest, on purpose. Clearing the blocks makes you eligible on all eight platforms. Eligible is not present, and no amount of money spent with us or anyone else converts one into the other. What we can promise is that the switches will be in the right position, that you will know which choices were yours, and that you will see the same numbers we do.
| What we found | On what evidence | Why it is not among the fifteen |
|---|
One of these is worth reading twice. Your review stream has been quiet since February. Professional opinion in our field says that matters; the published documentation does not say so. We record it, and we do not put it on your list — because a recommendation that rests on opinion rather than a rule is not something we are willing to charge you to act on. If you want reviews coming in again, that is a good idea for ordinary business reasons, and we will help. It is just not a finding.
There is no overall figure on this page on purpose. A single number would be ours, invented, and would tell you nothing you could act on — and the fifteen items are what you can act on.
Items 1, 3, 6, 8, 10 and 11 — the profile name, the footer block, the AI answers setting, the address, the phone number and the hours. All six are corrections to things that are wrong today, none needs anything from you, and together they are about a day of work.
Items 12 and 13, the two targeting jobs. Then item 7's crawler directives once you have told us what you want on the two training questions. Item 5's link removals run in the background across several weeks because they depend on other people replying.
Items 4 and 15 — consolidating the city pages and putting real evidence of your work on the service pages. These are the two that change the site rather than correct it, and they are the two that need you.
Photographs of three recent jobs, with the name of the technician who did each. Phone photos are fine — better than fine, actually.
Ten minutes on the phone about your profile categories, so we are removing the right two and keeping the right four.
Item 2 — the three staff reviews. We have identified them and we are not acting yet. Our review-process work is with a lawyer for review before it goes to any client, and we would rather tell you that plainly than put words in your mouth that we have not had checked. We will write to you the day it clears.
Our pre-ship checklist runs on every deliverable before it leaves. Nothing goes out with a fail on any item. P marks a prohibited practice; S marks something we may never say. Here are all twenty-four, in full, because a checklist nobody can read is a claim rather than a control.