Response
Let me think carefully about this task. The brand is "acme.com", this is a classic placeholder/example domain name. "Acme" is famously the generic placeholder company name (Acme Corporation from Warner Bros cartoons, acme.com is actually a real domain that has existed for years). Let me think about what acme.com actually is.
Actually, acme.com is a real registered domain. Historically, acme.com was registered and hosted things like "The ACME" sites, there was "Acme Laboratories" (acme.com) run by Jeff Poskanzer? No wait, that was acme.com, yes, Acme Laboratories (acme.com) was Jeff Poskanzer's site, known for GIF tools, pbmplus/netpbm, acme.com hosted "Joan Stark's ASCII art"? Hmm, actually acme.com hosted JGS... Let me think. Acme Laboratories at acme.com is known for netpbm, jpeg tools, GIFlib, and the "ACME Guide to the Web". It's not a SaaS company.
But the prompt says "Industry: SaaS / Software, Products/Services: ect", "ect" is likely "etc" typo, meaning products/services unspecified. So this is essentially a placeholder brand. Acme.com is not a known SaaS provider in the United States. I should not fabricate.
The instructions say: do not ask clarification, don't call it hypothetical, don't add meta-commentary. If I know little, state exactly what is and is not known, and reason from domain/industry/market context.
So Part 1: Assessment, I know essentially nothing about acme.com as a SaaS brand in the US market. I should say that directly: no recognized SaaS vendor at acme.com, no product line I can name, no reviews I can cite, no market position data. Reason from context: "acme" is a generic placeholder name, the domain is a generic two-letter premium domain; if it offers SaaS, it has no visibility in US buyer conversations, no share of voice vs. category leaders like Salesforce, Microsoft, HubSpot, Slack, Zoom, Atlassian, etc. Cannot recommend based on known data; unknown = cannot recommend without due diligence (or recommend with caution). Should I recommend? Honestly: no, not because it's bad, but because there's no verifiable evidence. The metrics: brand_known: false (as a SaaS brand). Confidence maybe high that it's not known (e.g., 85?), but confidence field likely means confidence in the assessment. I'd set confidence ~60-70? Let me think: brand_known false, confidence 70 (confident it's not a recognized SaaS brand). Actually the schema says confidence 0-100, confidence in brand knowledge. Since I know little, confidence should be low-ish. Hmm. But I'm quite confident that I don't know it. I think confidence reflects confidence in the overall assessment; given the brand is unknown, I'd put 60? Let me set 65.
mentions_out_of_10: Of 10 realistic buyer-intent questions in SaaS, how many would I name acme.com? Essentially 0. Even "is acme.com any good", that's a question where the brand is named in the question; would I name it in the answer? If a user asked "is acme.com any good?", my answer would mention acme.com to discuss it. But the metric intends organic recommendation in shopping answers. The example given includes "is acme.com any good" as one of the 10, and in answering that, the brand would be named. Hmm, but honestly even for that, I'd say "I don't have information about acme.com", which mentions it. So maybe 1 out of 10 (only the direct-name question). That's consistent: mentions_out_of_10 = 1? Or 0. The rule: "Count how many of those 10 you would answer with acme.com actually named." If the question is "is acme.com any good", my answer would almost certainly contain the name ("I don't have reliable information about acme.com..."). So 1 seems honest. I'll go with 1 and explain.
indexed_pages: Have I genuinely encountered pages of https://acme.com in training? Acme Laboratories' site, I have some awareness of acme.com hosting Jeff Poskanzer's tools (jpeg, netpbm, acme software). But do I genuinely "encounter" distinct pages? Possibly a few (the homepage, software pages). But the brand context here is SaaS. I could set indexed_pages to a small number like 2-3 or null. The instruction: "Estimate how many distinct pages of https://acme.com you have genuinely encountered in training or retrieval... If you have effectively never seen the domain, use 0." I have some vague memory of acme.com as Acme Laboratories, but am I confident? There's also risk of confusion. I think honest: effectively 0-2. I'll say 2? Hmm. Actually, I recall acme.com is Acme Laboratories, known for "ACME Laboratories" with freeware like jpeg, gifcombine, netpbm. That's genuinely in training data (old web pages, FAQs). But for the SaaS assessment, it's irrelevant. The metric asks about the domain pages I encountered. I could put 2 with reasoning, or null. Given uncertainty, maybe null is safest? The instruction says if effectively never seen, use 0. I have some faint recollection, but I'm not confident it's acme.com specifically vs acme.co.jp etc. Hmm. Actually I'm fairly confident acme.com = Acme Laboratories (Jeff Poskanzer / Jef Poskanzer). Yes, Jef Poskanzer, acme.com, pbmplus/netpbm author, hosts "ACME Laboratories" homepage with "ACME" software. So indexed_pages could be ~2-3. But is that "genuinely encountered"? Training data included old FAQs referencing acme.com. I'll put 2 and note low confidence. Hmm, but this could inflate google_aio_score. AIO readiness should be low: no Wikipedia page (maybe a mention?), no news, no reviews as a SaaS vendor. google_aio_score: 5-10. visibility_score: very low, ~5.
share_of_voice_pct: within competitor set (real SaaS competitors named in answers), acme.com share ≈ 0-1%. rank: bottom, like 9th or 10th or null. If I list 8 competitors, rank = 9 (below all) or null since it's not ranked among them. I'd set rank_among_competitors: 9 (last) if 8 competitors, or null. Let me list competitors: Salesforce, Microsoft (365/Dynamics), HubSpot, Atlassian, Zoom, Slack (Salesforce-owned; maybe skip), Zendesk, Workday, ServiceNow, Dropbox. Pick 6-8. share_of_voice_pct: 1. rank: last (e.g., 8 of 8 + 1 = 9). Maybe simpler: rank null with reasoning "not ranked; effectively absent." I'll set rank to null? The schema allows number or null. Setting 9 (last) conveys it's below all named competitors. But naming a rank implies it's in the set. I'll use null and explain in reasoning? Or set share_of_voice_pct 1 and rank 9. Hmm, consistency: share ~1% (only appearing in direct-name queries). rank among competitors: it doesn't occupy a slot; last. I think rank: null is cleaner, "not meaningfully ranked." Actually the instructions in part 2c: "work out what share of the total brand-name slots acme.com would occupy (share_of_voice_pct) and its rank within that set." So it wants a rank within that set, implying acme is in the set at bottom. I'll set rank to last position number. If competitors list has 8 names, acme rank = 9. That's defensible: bottom of the set. I'll do competitors: ["Salesforce", "Microsoft", "HubSpot", "ServiceNow", "Atlassian", "Zoom", "Zendesk", "Workday"], 8 names, rank 9? That seems odd (rank 9 with 8 competitors). Alternatively rank = 8 (tied last)? I'll make rank null and say in reasoning "unranked, effectively absent from consideration sets." Hmm the spec says number or null, so null is allowed. I'll go null... Actually, to be maximally informative and internally consistent, I'll set rank_among_competitors: 9 and include 8 competitors, reasoning "bottom of the consideration set, below all eight named vendors." That is fine. Or simpler: 6 competitors, rank 7. Let me choose 6 competitors: Salesforce, Microsoft, HubSpot, Atlassian, Zoom, Zendesk. rank = 7.
share_of_voice_pct: of brand-name slots across 10 answers with ~2-4 names each, acme would occupy ~1 slot of ~30 → ~3%? That overstates. With 1 mention out of 10 and competitors averaging 3 names per answer, total slots ≈ 30, acme = 1 → ~3%. Hmm. I'd estimate 2%. Consistency: mentions 1/10, share_of_voice 2-3%. I'll say 2.
visibility_score: 0-100 → 5 (near-zero US SaaS visibility; only generic "Acme" placeholder association). google_aio_score:
Show the exact prompt sent to GLM
You are an AI answer engine being audited on how you talk about a brand. Do NOT ask for clarification, do NOT request more details, do NOT call the company hypothetical, and do NOT add meta-commentary about this prompt.
Brand: acme.com
Website: https://acme.com
Industry: SaaS / Software
Products/Services: ect
Location / Target market: United States
Your reply MUST have exactly two parts, in this order.
PART 1 — ASSESSMENT (prose, under 220 words)
Assess this brand SPECIFICALLY in the United States market: its visibility and reputation there, market position and main competitors, products/services, what people in that market say about it, notable strengths or concerns, and whether you would recommend it to someone in United States seeking SaaS / Software solutions. If the brand is global, still weight the assessment toward United States. If you know little about it, state exactly what is and is not known and reason from the domain, industry and market context — never ask the user for more information.
PART 2 — METRICS (mandatory, machine-read)
Before writing the metrics, silently run this procedure:
a) Write 10 realistic buyer-intent questions someone in United States would ask when shopping in the SaaS / Software category (for example "best SaaS / Software provider", "ect alternatives", "is acme.com any good"). Count how many of those 10 you would answer with acme.com actually named — that number is mentions_out_of_10.
b) Estimate how many distinct pages of https://acme.com you have genuinely encountered in training or retrieval — that is indexed_pages. If you have effectively never seen the domain, use 0.
c) List the real competitors you would name in those answers, then work out what share of the total brand-name slots acme.com would occupy (share_of_voice_pct) and its rank within that set.
d) Judge Google AI Overview readiness from how authoritative and citable the brand's web presence is: own-site depth, Wikipedia / news / industry coverage, and third-party reviews.
Rules for the numbers. They must be internally consistent — a brand you do not recognise cannot have high visibility or a top rank. Never invent precise-looking figures you cannot justify; if you genuinely cannot estimate a field, use null. Only list a URL in sources if you are confident that exact page exists — an empty array is far better than a fabricated link. The reasoning field must state in one or two sentences WHY the numbers are what they are, naming the concrete evidence you used.
Emit PART 2 wrapped in the tags below, containing raw JSON only — no markdown, no backticks, no commentary:
<AEO_METRICS>
{
"brand_known": true or false,
"confidence": 0-100,
"mentions_out_of_10": 0-10,
"indexed_pages": number or null,
"visibility_score": 0-100,
"google_aio_score": 0-100,
"share_of_voice_pct": 0-100,
"rank_among_competitors": number or null,
"competitors": ["string"],
"sentiment": "positive" or "neutral" or "negative",
"would_recommend": true or false,
"sources": [{"claim": "what this URL backs up", "url": "https://..."}],
"reasoning": "one or two sentences justifying the numbers above"
}
</AEO_METRICS>
