We Measured 7 of Our Own Websites in ChatGPT, Perplexity and Google AI Mode
By the Leadee team at Netbase JSC · Updated
Netbase JSC ran its own product, Leadee, against 7 of its own websites: only 2 of 7 brands (CMSmart and Printcart) are recommended by AI engines when a buyer has not already typed the brand name. The other 5 scored zero in 37-42 unbranded runs each. AI engines build their supplier shortlists from third-party sites — clutch.co, apps.shopify.com, G2, Capterra — not from the brands' own websites. All 7 sites scored A or B on technical GEO readiness, so technical readiness was not the bottleneck.
About this measurement
Netbase JSC owns all 7 sites measured here (CMSmart, Printcart, SoftWSP, NetbaseJSC, Web2Print, AdminKit, OVN/OutsourcingVN). This page is Netbase measuring its own portfolio with its own product, not an independent audit.
This is a baseline measurement, not a result. Netbase JSC's own sites have not shipped the fix plan below yet — the earliest honest re-measurement is about four weeks after each site does, using the same prompts and engines. Nothing on this page claims an improvement.
- Period
- 2026-W43
- Change
- CHG-20261025-LDE01
- Sites measured
- 7
- Perplexity sonar: n=3 per prompt, full sample across all 7 sites
- Google AI Mode (via DataForSEO): n=3 per prompt, full sample across all 7 sites
- OpenAI gpt-5 + web search: only 25 runs total across all 7 sites — the OpenAI account ran out of credit (insufficient_quota) partway through
Headline numbers
- Brands recommended, unbranded questions
- 2 of 7
- Mentions outside the US, all 7 brands
- 4
- Sites scoring A or B on GEO audit
- 7 of 7
- Actual cost of this measurement round
- $3.95
What we found
- Only 2 of 7 brands were recommended by AI to a buyer who had not already named them: CMSmart (32.5% of unbranded US questions) and Printcart (17.5%). The other 5 brands — NetbaseJSC, Web2Print, SoftWSP, OVN, AdminKit — scored 0% across 37-42 unbranded runs each. AI only named these 5 when the question already contained the brand name.
- The second market is nearly blank: UK, Germany and Singapore combined produced 4 mentions across all 7 brands (CMSmart 2/22 UK, Printcart 1/14 UK, every other brand 0).
- AI engines build their shortlists from third-party sites, not from the brand's own site: clutch.co (B2B services), apps.shopify.com (Shopify app customizers), designnbuy.com/G2/Capterra (web-to-print), mxtoolbox.com (DNS/email tools), binarly.io/finitestate.io (firmware security). The two winning brands are also the two whose own domain is cited most in their niche (cmsmart.net cited 96 times, printcart.com + solution.printcart.com cited 87 times combined).
- Technical readiness was not the main bottleneck: all 7 sites let AI crawlers in, and every one scored an A or B on Leadee's core GEO audit. The recurring technical issues were narrower: missing schema.org markup (Printcart, SoftWSP, Web2Print), slow TTFB (OVN 2.5s, Printcart 1.9s, Web2Print 1.6s), llms.txt format errors (6 of 7 sites) and pages with no visible update date (Printcart 11 of 15 pages, SoftWSP 7 of 15).
- Ambiguous brand names dilute the entity AI engines try to resolve: "Web2Print" overlaps a generic industry term, "AdminKit" collides with the unrelated Bootstrap template adminkit.io, "OVN" is a non-obvious acronym, and Printcart's own citations are split between printcart.com and solution.printcart.com.
All 7 sites, side by side
"Mentioned" counts an unambiguous form (full name/domain, or a citation). 95% confidence intervals are Wilson. Engines: Perplexity sonar and Google AI Mode ran the full sample (n=3); OpenAI gpt-5 reached only 25 runs total (all 7 brands combined) because the account ran out of credit — see "Limits of this measurement".
| Brand | Mentioned, all US questions | Mentioned, unbranded US questions | SoV US | Second market | Leading competitor (US mentions) | GEO audit | Citability (avg.) |
|---|---|---|---|---|---|---|---|
| CMSmart | 46% (24/52; CI 33–60%) | 33% (13/40) | 17% | GB 2/22 | Fancy Product Designer (39) | A (86.8) | 0.26 |
| Printcart | 37% (19/52; CI 25–50%) | 18% (7/40) | 15% | GB 1/14 | Zakeke (35) | B (79.6) | 0.26 |
| SoftWSP | 24% (12/50; CI 14–37%) | 0% (0/38) | 15% | DE 0/12 | Finite State (22) | A (86.7) | 0.24 |
| NetbaseJSC | 20% (11/54; CI 12–33%) | 0% (0/42) | 16% | GB 0/25 | FPT Software (19) | A (95.5) | 0.28 |
| Web2Print | 14% (7/50; CI 7–26%) | 0% (0/38) | 9% | GB 0/15 | DesignNBuy (26) | B (80.6) | 0.27 |
| AdminKit | 10% (5/51; CI 4–21%) | 0% (0/39) | 10% | GB 0/24 | MXToolbox (23) | A (86.2) | 0.26 |
| OVN | 8% (4/49; CI 3–19%) | 0% (0/37) | 8% | SG 0/22 | Rikkeisoft (14) | A (92.4) | 0.26 |
What this measurement cost
Actual cost of this measurement round: $3.95 against a round budget of $30 (platform cap $300/month; $4.23 used so far this month).
- OpenAI gpt-5 + web search: 25 runs = $2.39 (≈ $0.096/run)
- Perplexity sonar: 252 runs = $1.30 (≈ $0.005/run)
- Google AI Mode (DataForSEO): 215 runs = $0.26 (≈ $0.0012/run)
- Audit, content audit, Prompt/Demand Discovery, Fix Engine: $0
Limits of this measurement
- OpenAI ran out of credit (429 insufficient_quota) after 25 runs total across all 7 brands.
- DataForSEO used up its $1 trial credit (402), which limited Google AI Mode in the second market.
- Gemini was not measured (the Google AI Studio project has no billing enabled yet).
- Claude was not measured (no API key configured yet).
- A follow-up round will add the missing OpenAI runs (about $22 for all 7 brands) and the missing second-market Google AI Mode runs once credit is topped up.
What we will change, in priority order
Get listed where AI builds its shortlists
nowClutch, GoodFirms, DesignRush (services), G2, Capterra, GetApp (software), Shopify App Store (apps) — applies to NetbaseJSC, OVN, Web2Print, Printcart, AdminKit. 5 of 7 brands scored 0% on unbranded questions, and third-party sources decide the shortlist.
Publish honest "X vs leading competitor" and "alternatives to" pages
nowApplies to all 7 brands. Comparison and alternatives prompts currently return the competitor only.
Add Organization + Product/Service/SoftwareApplication schema.org with sameAs
nowPrintcart, SoftWSP, Web2Print, AdminKit, OVN are missing schema; use one full, consistent entity name per brand.
Open every service/pricing page with an answer-first block: numbers + an update date
nowAll 7 brands: content-audit citability averages only 0.24–0.28. Pages are missing answer-first paragraphs, specific numbers and a visible source.
Cut TTFB below 800ms
nowOVN, Printcart, Web2Print. AI crawlers skip slow pages.
Build content for the second market (UK/DE/SG) that cites local sources
nextAll 7 brands: 4 mentions across all 7 brands outside the US. Local sources like sortlist.co.uk, capterra.co.uk, bsi.bund.de, ncsc.gov.uk get cited there.
Merge solution.printcart.com into printcart.com; refresh CMSmart's 2020-2021 content
nextPrintcart, CMSmart. These are the 2 winning brands — protect what is working before fixing what is not.
Questions
Why does Netbase publish a case study where 5 of its own 7 sites score zero?
Because the honesty rule applies to Netbase's own sites too (spec seo-geo-cluster-plan.md §0): no invented numbers, and this is what a real, unflattering baseline looks like before any fix ships.
Is this an independent audit of these 7 brands?
No. Netbase JSC owns all 7 sites measured, and this measurement was run by Netbase using its own Leadee product. See "Conflict of interest" above.
Why only 25 OpenAI runs out of hundreds planned?
The OpenAI account used for this round ran out of credit partway through (429 insufficient_quota). The 25 runs that did complete are included and labelled; the rest of this baseline relies on Perplexity sonar and Google AI Mode, both run to the full planned sample.
When will these 7 sites be re-measured?
About four weeks after each site ships its own fix plan, using the same prompt sets and engines, so any change is statistically meaningful rather than noise.
Can GEO guarantee results?
No. AI answer engines change their output frequently, and no vendor — including Leadee — can guarantee a ranking or a mention. Leadee publishes confidence intervals on every metric so you can see how much a result can vary run to run.
How does Leadee measure AI visibility?
Leadee runs your prompt set against each AI engine on a weekly to daily schedule, depending on plan, through official vendor APIs (OpenAI, Gemini, Perplexity, Anthropic, DataForSEO) — never by scraping consumer chat interfaces — and reports mention rate, citation rate, average position, share of voice and sentiment with a 95% confidence interval.
Who do I talk to?
Email [email protected] or use Talk to an expert at https://leadee.ai/contact-sales — the Netbase JSC team behind Leadee replies by email or WhatsApp — or WhatsApp the CEO directly at https://wa.me/84937869689.
See where your own site stands
Run a free AI visibility scan on your own domain — the same kind of check Leadee ran on Netbase JSC's own 7 sites.