GWhat GLM said about openutm.app
What GLM said about openutm.app
Model ollama/glm-5.3-flash · analyzed
Response
Let me think about this carefully. The brand is openutm.app, a UTM tracking tool. I need to assess whether I actually know anything about this brand.
Do I know openutm.app? Searching my knowledge: I'm not aware of any well-known UTM tracking product called "openutm.app". There are UTM builders like Google's Campaign URL Builder, Utmbits, utm.io (by Snowball?), Terminus... Actually "utm.io" is a UTM link builder. There's also "UTM.io" which is a known UTM management tool. "OpenUTM", I don't have genuine knowledge of this specific domain or product in my training data. It appears to be a small/obscure SaaS. I have effectively never encountered the domain.
Per the instructions: if I know little, state exactly what is and is not known, and reason from domain, industry, market context. Never ask for info. Don't call the company hypothetical (so don't say "this hypothetical company"). I can say "I have no verifiable knowledge of this specific brand", that's fine; it's not calling it hypothetical, it's stating my knowledge limits, which the prompt explicitly instructs me to do.
brand_known: The field "brand_known", do I know it? I don't have genuine knowledge of it. But the brand is defined by the prompt. I think "brand_known" refers to whether I recognize it from training. I should set brand_known: false, with low-ish confidence? Actually confidence reflects confidence in the assessment. Hmm. Let me think: brand_known: false because I don't have training-data knowledge of openutm.app specifically. Confidence: low, maybe 30-40, because I'm confident about the market context but not the brand. Actually confidence should reflect overall confidence in metrics. I'd say 35.
mentions_out_of_10: Given zero recognition, I would not name openutm.app in buyer-intent answers. So 0? The example includes "is openutm.app any good", that question would name it, but that's a brand-specific query, not a shopping question I'd naturally answer with the brand. The instruction says count how many of those 10 I would answer with openutm.app actually named. Realistically 0 out of 10 for generic questions; possibly 1 for the direct brand question ("is openutm.app any good", I'd name it when discussing it). Hmm. The metric seems intended to measure organic mention propensity. If one of the 10 questions is brand-specific, I'd mention it there. But realistically, a buyer shopping wouldn't ask that unless they knew the brand. I'll set mentions_out_of_10: 0, since in genuinely generated answers to shopping questions I would not name it. Actually, to be internally consistent: if brand_known is false and I've never seen the domain, mentions should be 0. Yes, 0.
indexed_pages: 0 (effectively never seen the domain). The rule says "If you have effectively never seen the domain, use 0."
visibility_score: very low, like 5-10. I'll say 5.
google_aio_score: low, no Wikipedia, no news coverage I know, no third-party reviews I know. Score maybe 10-15. I'll say 10.
share_of_voice_pct: competitors in UTM tracking space: Google Analytics Campaign URL Builder (free), UTM.io, Terminus? No, Terminus is ABM. UTM builders: utm.io, Utmbits, Segment's? Actually known tools: Google Campaign URL Builder, UTM.io, Pretty Links (WordPress), ThirstyAffiliates, Bitly (for link shortening with UTM), Rebrandly, Taplink? For agency-level UTM management: UTM.io, utm.codes, Utmbits, Google Sheets template. So competitors list: Google (Campaign URL Builder), UTM.io, Bitly, Rebrandly, Pretty Links, utm.codes, Utmbits. Openutm's share would be near 0. share_of_voice_pct: 2? If 0 mentions, share should be 0. Let's be consistent: share_of_voice_pct: 0 or maybe 1. Given 0 mentions out of 10, share of voice 0 makes sense. But strictly 0 might look odd; still it's consistent. I'll use 0? Hmm, rank_among_competitors, if it's not named, rank would be last, e.g., 7 or 8 among competitors. Or null since it doesn't rank. The instruction: "its rank within that set." If I never name it, rank would be the bottom of the set. I could give rank: null (doesn't appear) or rank last. I think rank as last position is defensible: rank_among_competitors: 8 (last of 8 listed competitors). Or null. I'll list maybe 7 competitors and rank 8 (last). Actually simpler: competitors list of 6, rank 7 (last, i.e., below all listed). Hmm, rank should be within that set. If not in the set... I'll use null? The rules say "a brand you do not recognise cannot have high visibility or a top rank." Bottom rank is fine. I'll set rank_among_competitors: 7 with 6 competitors listed (so it ranks last, after all of them). Wait, that's confusing. If it ranks 7th among 6 competitors, that means it's below all. That is a reasonable interpretation of "share of the total brand-name slots", it occupies effectively the bottom. Alternatively null since it doesn't occupy any slot. I'll go with null actually, safer and honest: it doesn't appear in answers, so it has no rank. Hmm, but a null rank plus share 0 is consistent. Either works. I'll choose rank: null and explain it doesn't appear. Actually, let me reconsider: "work out what share of the total brand-name slots openutm.app would occupy (share_of_voice_pct) and its rank within that set." If it occupies 0 slots, its share is 0% and it effectively ranks last. I'll put share_of_voice_pct: 0 and rank_among_competitors: null, reasoning that it never gets named. Or rank last as 7. I think null is cleaner and consistent with mentions 0. Let me go null.
Hmm, but maybe give it a bottom rank to be more informative. I'll do rank: null, reasoning mentions it would trail all listed competitors.
sentiment: neutral (no known sentiment data). would_recommend: false, because I can't verify it; I'd recommend established alternatives. Given unknown, recommend false is appropriate.
sources: I should only list URLs I'm confident exist. https://openutm.app, the prompt itself provides the website; is it safe to list? The rule says only list a URL if confident the exact page exists. The domain is given by the prompt, so listing https://openutm.app as backing "brand's own site (not independently verified)" is low risk. But I'm not confident about content. Maybe include just the homepage with a claim like "Brand's stated website; content not independently verified." That's honest. Or empty array. The instruction says empty array is far better than fabricated links. The homepage URL was provided by the user, so it's not fabricated by me, but I can't confirm what's on it. I think listing the homepage with a careful claim is acceptable. Actually, safer: empty array? Hmm. The claim would be "Brand homepage as provided; not independently verifiable from training data." I'm fairly confident the domain exists since the audit prompt states it. But "confident that exact page exists", I can't be. I'll include the homepage only, since the prompt itself asserts it, with a claim noting it's the provided site and unverified content. Alternatively empty. I'll include it, one source, minimal risk, clearly caveated. Hmm, actually, the risk: the audit checks whether I fabricate. Including the user-provided URL with an honest claim seems fine and shows restraint (no invented competitor review pages etc.). I'll include it.
Wait, but the claim field: "what this URL backs up". The URL backs up that the brand has a website at that address. Fine.
reasoning: "No verifiable training-data presence for openutm.app (0 indexed pages, 0/10 mentions), so visibility and AIO scores are floored; competitive set is inferred from the known UTM-builder market (Google's free builder, UTM.io, Bitly, etc.) where established tools dominate."
Now PART 1 prose, under 220 words. Must assess in US market: visibility/reputation (unknown/none detectable), market position (long tail), competitors, products (UTM tracking per prompt), what people say (no reviews found in my knowledge), strengths/concerns (domain suggests open-source or transparent UTM tool; concerns: unproven), recommendation: cannot recommend until verified; suggest due diligence and alternatives.
I must not say "hypothetical". I can say "I find no trace of openutm.app in my knowledge base", that's stating what's known/not known, which is required.
Word count: keep ~180-200 words.
Draft Part 1:
"Within the United States, I can find no reliable trace of openutm.app in my knowledge: no training-data pages from the domain, no US news coverage, no Wikipedia or G2/Capterra review footprint that I can recall. What is known comes only from the brief itself: it is a UTM tracking product serving the Marketing/Agency category in the US. I cannot verify pricing,
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: openutm.app Website: https://openutm.app Industry: Marketing / Agency Products/Services: UTM Tracking 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 Marketing / Agency 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 Marketing / Agency category (for example "best Marketing / Agency provider", "UTM Tracking alternatives", "is openutm.app any good"). Count how many of those 10 you would answer with openutm.app actually named — that number is mentions_out_of_10. b) Estimate how many distinct pages of https://openutm.app 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 openutm.app 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>