Public sample

AI Visibility and Tech Audit for
GOA Studio

How London homeowners find residential architects — and don't find GOA Studio — across ChatGPT, Gemini, Perplexity, Claude, and Google AI Overviews. Audited across 10 buyer prompts, 5 AI engines, and 5 site pages. Looking at your Crawlability, Content Optimisation, Sentiment, and Rankings.

Default visibility: public. Anyone with the link can read this report. Sign in to your RankBee account to make it private to your team.
GS
GOA Studio
goastudio.co.uk
GeneratedMay 11, 2026
Audit windowLast 14 days
Report IDRB-2026-05-11-0001
What's in this report

Four sections covering technical access, AI visibility, content, and reputation.

This is more than a crawl audit. We measure where your buyers go to find you, what AI says when they ask, and what's missing from your story.

01Content Scorecard

Content quality: How well does your content perform vs the competition?

RankBee crawled 5 GOA Studio pages and scored each 1–10 against live competitor URLs across relevant buyer queries. GOAStudio ranks #1 on local and practice pages but scores below 2.0 on editorial and discovery pages — the gap that explains the AI citation absence.

Page-by-page scoring
As % · 5 pages graded
39% your avg39% leader avg
Page
Your score
Leader
Δ
Homepage
https://goastudio.co.uk
58%
18%
https://urbanistarchitecture.co.uk/house-extension-architects/
40%
Portfolio (filtered)
https://goastudio.co.uk/portfolio-filter
46%
22%
https://www.simplyextend.co.uk/
24%
Kitchen Extension Guide
https://goastudio.co.uk/media-features-publications/design-expert-advice-for-kitchen-extensions/
15%
23%
https://k-c-a.co.uk/blog/kitchen-extension-guide/
7%
Enfield Architect
https://goastudio.co.uk/enfield-residential-architect-projects/
57%
26%
https://www.lcclconstruction.co.uk/areas/enfield
30%
Portfolio Map
https://goastudio.co.uk/portfolio-map/
18%
36%
https://goastudio.co.uk/
18%

Content quality leaderboard

i
Weighted average across audited pages
Brand
GAIO Score
Avg Rank
1.
GOA Studio
39%
2.40
2.
k-c-a Architecture
23%
1.00
3.
Simply Extend
17%
3.00
4.
Urbanist Architecture
15%
2.00
5.
Architecture for London
13%
4.00
02AI Rankings Matrix

Visibility coverage: where you appear vs. competitors

Real buyer prompts, run live against 5 AI engines — ChatGPT (GPT-5.4), Gemini, Perplexity (Sonar), Claude (Sonnet 4.5), and Google AI Overviews. Brand mentions counted when GOAStudio or a competitor is named in the AI answer or cited as a source URL. 50-cell matrix: 10 prompts ×...

ChatGPT
GPT-5.4
10%you
vs 20% Urbanist Architecture · -10 pp gap
Gemini
3.1 Flash Lite
20%you
vs 10% Urbanist Architecture · +10 pp gap
Perplexity
Sonar
10%you
vs 10% Urbanist Architecture · +0 pp gap
Claude
Sonnet 4.5
20%you
vs 0% Urbanist Architecture · +20 pp gap
AI Overviews
Google AIO
20%you
vs 0% Urbanist Architecture · +20 pp gap
AI coverage matrix
All 10 prompts shown
YouUrbanist Architecture (leader)Scenario ArchitectureExtension ArchitectureShape Architecture
#
Prompt
ChatGPT
Gemini
Perplexity
Claude
AI Overviews
1
Vendor Evaluation
Best residential architect in London for house extensions 2026
2
Vendor Evaluation
GOAStudio vs Shape Architecture vs Scenario Architecture for London home extensions
3
Vendor Evaluation
Top RIBA architects for conservation area extensions in London
4
Operational
How does the planning process work for a house extension in London
5
Operational
What are the risks of a rear house extension in a conservation area London
6
Operational
How long does it take to get planning permission for a home extension in London
7
Compliance & Regulation
What building regulations apply to a kitchen extension in London
8
Compliance & Regulation
Do I need planning permission for a single storey extension in London
9
Services & Costs
How much does a residential architect cost for a house extension in London
10
Services & Costs
What architectural services do I need for a loft conversion in London

AI Coverage Leaderboard

i
Across all 10 prompt × 5 model cells
Brand
GAIO Score
Avg Rank
1.
UA
Urbanist Architecture
7%
2.00
2.
GS
GOA Studio
7%
3.00
3.
SA
Scenario Architecture
4%
3.00
4.
EA
Extension Architecture
4%
4.00
5.
SH
Shape Architecture
0%
0.00
03AI Crawlability Audit

Identify risk areas for your AI SEO Crawl strategy

Before your content can be cited, it has to be crawled and read. We tested five layers: 1) What your robots.txt declares, 2) what real users experience, 3) what AI search agents retrieve through WebSearch, 4) what bots get past your CDN, and 5) how deep into the rendered page...

PHASE 1

Robots.txt analysis

Permissive — all bots allowed

What your robots.txt declares to each AI crawler, and which bots are allowed, blocked, or partially restricted.

RiskLowCrawlers24Allowed24Blocked0Partial0
robots.txt200· 13 linesFetched 2026-05-11 15:22 UTC
🚨Key risks flagged
🛠
🔍
allowed!partialblocked
Bot
Provider
Role
Status
Rule applied
GPTBot
OpenAI
Training crawler
Allow
Explicitly Allow: / via user-agent rule
ChatGPT-User
OpenAI
User-triggered fetch
Allow
Inherits * — no block
OAI-SearchBot
OpenAI
Search indexing
Allow
Inherits *
ClaudeBot
Anthropic
Training crawler
Allow
Inherits *
Claude-User
Anthropic
User-triggered fetch
Allow
Inherits *
Claude-SearchBot
Anthropic
Search indexing
Allow
Inherits *
anthropic-ai
Anthropic (legacy)
Training crawler
Allow
Inherits *
Google-Extended
Google
Training opt-out flag
Allow
Inherits *
GoogleOther
Google
General fetcher
Allow
Inherits *
PerplexityBot
Perplexity
Search indexing
Allow
Inherits *
Perplexity-User
Perplexity
User-triggered fetch
Allow
Inherits *
CCBot
Common Crawl
Training crawler
Allow
Inherits *
Bytespider
ByteDance
Training crawler
Allow
Inherits *
Meta-ExternalAgent
Meta
Search indexing
Allow
Inherits *
Meta-ExternalFetcher
Meta
Content fetcher
Allow
Inherits *
Applebot-Extended
Apple
Search indexing
Allow
Inherits *
Amazonbot
Amazon
Search indexing
Allow
Inherits *
DuckAssistBot
DuckDuckGo
Search indexing
Allow
Inherits *
Diffbot
Diffbot
Search indexing
Allow
Inherits *
Omgilibot
Webz.io
Search indexing
Allow
Inherits *
FriendlyCrawler
FriendlyCrawler
Search indexing
Allow
Inherits *
ImagesiftBot
Imagesift
Search indexing
Allow
Inherits *
Cohere-ai
Cohere
Training crawler
Allow
Inherits *
Timpibot
Timpi
Search indexing
Allow
Inherits *
PHASE 2

Virtual user crawl test

2 probes — 200 OK

Headless visits from a 🇬🇧 GB IP and a 🇺🇸 US IP confirm the site is reachable for real readers — and therefore reachable for AI crawlers that proxy through the same regions. This is a sanity check, not a deep audit.

🇬🇧GBsuccess
✅ Accessible from GB IP — 200 OK. Site loads correctly for local UK users and crawlers.
200 HTTPblocked: false
🇺🇸USsuccess
✅ Accessible from US IP — re-tested 2026-05-11 via three independent US-origin crawls (GPTBot, ClaudeBot, Googlebot). All returned HTTP 200 with full page content. The original audit 408 was a one-off cold-connection probe timeout from a different infrastructure path, not an active geo-block or WAF rule. Response times from US range 10–26s — slow but not blocked. CDN edge coverage (Cloudflare) recommended to reduce US response times.
200 HTTPblocked: false
What this test returns6 fields per country
{
  "countryCode": "US",
  "status":      "success",
  "blocked":     false,
  "statusCode":  200,
  "error":       "",
  "summary":     "✅ Accessible from US IP"
}
The 6 fields
countryCodeISO 3166-1 alpha-2 country the test ran from
statusHigh-level outcome: success / failed / error
blockedWhether the site rejected the visitor (geo or anti-bot)
statusCodeHTTP status from the origin (e.g. 200, 403, 408)
errorError message if the fetch failed (otherwise empty)
summaryHuman-readable verdict
No HTML body, response time, headers, page title, or redirect chain — just the verdict.
PHASE 3

LLM web-search access

4 of 4 reachable

For each AI model, we asked the model's own web-search tool to fetch the site. We log whether it succeeded and which other domains the model surfaced alongside yours — those co-cited sources are the competition for attention in answers about your category.

Provider
Model
Status
Co-cited sources
Notes
OpenAI (GPT-5.4)
gpt-5.4
Reachable
urbanistarchitecture.co.ukresi.co.ukextensionarchitecture.co.uk
Accessible — used web search and returned accurate homepage content ('Designing homes around the way you live') but did not cite goastudio.co.uk as a source URL. GPT retrieved and read the brand's content but resolved citations from third-party directories and aggregators instead. This is a citation behaviour issue, not an access issue: owned content is being read but not linked. Third-party citations dominate AI answers for extension queries because GOAStudio lacks the dedicated planning and compliance content that earns direct citations.
Anthropic (Claude Sonnet 4.6)
claude-sonnet-4-6
Reachable
resi.co.ukarchitectureforlondon.comhouzz.co.uk
Accessible — fetched and accurately summarised the GOAStudio homepage (George Omalianakis, RIBA, conservation areas) but returned zero citations pointing to goastudio.co.uk. Claude's web search retrieved content but resolved citations from aggregator and directory pages instead of the brand's own URLs. Same citation behaviour pattern as OpenAI: the site is readable, but owned pages are not being cited because the content doesn't match what buyers are searching for — planning, compliance, and cost guidance.
Gemini (Flash Lite Preview)
gemini-3.1-flash-lite-preview
Reachable
goastudio.co.uk
Fully accessible — Gemini cited goastudio.co.uk directly and returned accurate brand summary. Google's infrastructure (GB-based) avoids the US timeout issue. This is GOAStudio's strongest AI channel at query-time and should be the anchor for content optimisation targeting Google AI Overviews.
Perplexity (Sonar)
sonar
Reachable
goastudio.co.ukhouzz.co.ukfind-and-update.company-information.service.gov.ukconstructionlondon.co.uk
Accessible — Perplexity cited goastudio.co.uk but also heavily pulled third-party directory and company registration sources (Houzz, Companies House, ConstructionLondon). The brand's own inner pages (service pages, project pages) are not being cited — only the homepage. Expanding rich inner-page content will push owned URLs further into Perplexity answers.
PHASE 4

Bot impersonation test

5 slow

We sent requests using each bot's exact User-Agent string. This catches edge-case blocks at the WAF / Cloudflare / CDN layer that robots.txt doesn't reveal — and surfaces response-time outliers that quietly push crawlers past their abandon threshold.

Bot
Status
HTTP
Response time
oai-searchbot
accessible
200
3,900ms
chatgpt-user
accessible
200
2,000ms
gptbot
accessible
200
16,000ms
chatgpt-agent
accessible
200
1,900ms
perplexitybot
accessible
200
1,800ms
perplexity-user
accessible
200
2,100ms
googlebot
accessible
200
7,800ms
googlebot-smartphone
accessible
200
12,527ms
bingbot
accessible
200
14,200ms
bing-copilot
accessible
200
2,100ms
claudebot
accessible
200
31,100ms
claude-user
accessible
200
1,800ms
claude-searchbot
accessible
200
36,900ms
grok
accessible
200
2,100ms
deepseek
accessible
200
1,500ms
Two patterns to fix: (1) bingbot times out where every other bot succeeds — likely a CDN rule treating the legacy Bing user-agent as suspicious, while the modern bing-copilot sails through. (2) Slow bots (10s+) respond, but their response times are 5–10× the typical ~2s baseline. Most LLM crawlers abandon at 3–5s, so those bots are likely truncating or skipping your pages even when the HTTP says 200.
PHASE 5

Indexability · token depth

5 of 5 pages at risk

Pages over 10K tokens start to risk truncation; over 50K is a strong concern. Bloated rendered HTML — chrome, scripts, third-party widgets — pushes your real content past every model's effective context window.

Page
10K50K100K
Tokens
Status
Homepage
https://goastudio.co.uk
18.0K
At risk
Portfolio (filtered)
https://goastudio.co.uk/portfolio-filter
16.0K
At risk
Kitchen Extension Guide
https://goastudio.co.uk/media-features-publications/design-expert-advice-for-kitchen-extensions/
20.0K
At risk
Enfield Architect
https://goastudio.co.uk/enfield-residential-architect-projects/
17.0K
At risk
Portfolio Map
https://goastudio.co.uk/portfolio-map/
15.0K
At risk
Why these pages are heavy5 explanations
Homepage · https://goastudio.co.uk
Estimated ~18,000 tokens total. SiteGround's optimizer inlines a combined CSS bundle directly in the <head> as both a <link> and a <style> block, plus Divi builder and Yoast Premium scripts — consuming an estimated 12,000–15,000 tokens before any body copy is reached. AI crawlers with token budgets (ChatGPT, ClaudeBot) will spend the majority of their context window on boilerplate before reaching the 'Designing homes around the way you live' headline. Fix: enable SiteGround's 'Combine CSS' to an external file reference only (no inline) and defer non-critical JS.
Portfolio (filtered) · https://goastudio.co.uk/portfolio-filter
Estimated ~16,000 tokens total, of which virtually none is indexable project content. The portfolio grid is rendered entirely by JavaScript — bots receive the same heavy <head> boilerplate as the homepage followed by an empty container div. No static project titles, descriptions, locations, or case study text exists in the served HTML. AI engines cannot index any project data from this page. Immediate fix: add a static HTML section below the JS filter listing all projects with title, borough, project type, and a one-sentence description.
Kitchen Extension Guide · https://goastudio.co.uk/media-features-publications/design-expert-advice-for-kitchen-extensions/
Estimated ~20,000 tokens total. Article body content is present in static HTML and indexable, but the same SiteGround head bloat (~12,000–15,000 tokens) precedes it. The article itself is thin — design tips only, no PD limits, no Building Regs, no Party Wall Act guidance, no cost ranges. This explains the 1.54/10 score (rank 4/16): AI engines find the page but don't cite it because the content doesn't answer planning or compliance queries. The token depth issue compounds the content quality problem — bots with tighter budgets may not reach the article body at all.
Enfield Architect · https://goastudio.co.uk/enfield-residential-architect-projects/
Estimated ~17,000 tokens total. Page body content (project descriptions, area copy) is present in static HTML and reaches AI engines — consistent with its strong 5.65/10 score (rank 1/16). Same head bloat pattern applies (~12,000–15,000 tokens before body), but the page is short enough that bots still reach the substantive copy within budget. Risk increases as the page grows. Recommend splitting into multiple borough sub-pages (Enfield Chase, Grange Park, Winchmore Hill) to keep per-page token counts manageable and add geo-specific content depth.
Portfolio Map · https://goastudio.co.uk/portfolio-map/
Estimated ~15,000 tokens total, of which virtually none is indexable project content. Like the portfolio filter, the map is rendered by a JavaScript mapping library (likely Leaflet or Google Maps) — all project pin data is loaded client-side from a JSON endpoint that bots cannot access. The static HTML body is essentially empty aside from a map container div. This is why GOAStudio's own homepage (3.62/10) outranks this page (1.84/10) on map-related queries — the homepage has more readable text than the map page itself. Fix: add a 'Projects by London Borough' static section beneath the map with borough headings and one project example each.
04Sentiment Snapshot

Brand perception: how AI models describe you to buyers

10 buyer prompts grouped into 4 intent clusters. Sentiment scored from live AI responses across ChatGPT, Gemini, Perplexity, Claude, and Google AI Overviews — 50 cells total. 'Positive' = GOAStudio mentioned/recommended. 'Absent' = brand not named by any engine on that prompt.

Operational / Planning
3 prompts · 12 model responses analysed
Absent

GOAStudio is completely absent from all 15 cells across the 3 operational prompts (planning process, conservation area risks, planning permission timelines). Engines instead cite Planning Portal, local authority websites (enfield.gov.uk), architecture advice guides (architectureforlondon.com), and general planning resources. No GOAStudio owned content appears as a cited source for planning advice. This is the most critical GAIO gap: the planning and process query cluster has the highest buyer intent frequency, and GOAStudio has zero presence. Publishing a comprehensive London planning guide (PD limits, LDC, Building Regs, timelines) under its own domain would directly address this 0/15 absence.

Compliance & Regulation
2 prompts · 8 model responses analysed
Absent

GOAStudio returns 0/10 cells across building regulations (P7) and planning permission (P8) queries. Extension Architecture appears in P7 (Gemini: 1 cell) and is the only architect brand cited in this cluster. The compliance cluster is dominated by government sources (planning.gov.uk, Building Regulations guidance), Checkatrade, and general building guides. GOAStudio's kitchen extension guide exists but scores 1.54/10 and does not contain the PD limits, Building Regs detail, or Party Wall Act guidance that would make AI engines cite it on these queries. Expanding the guide into a full compliance reference document would be the most direct route to earning citations in this cluster.

Services & Costs
2 prompts · 8 model responses analysed
Absent

GOAStudio absent from all 10 cells across architect cost (P9) and loft conversion services (P10) queries. For cost queries, engines cite HomeOwners Alliance, Checkatrade, and builder price-comparison guides. For loft conversion services, engines cite specialist loft firms (Scenario Architecture appears in P10 on Gemini). GOAStudio does offer both services but has no public pricing or cost-range guidance on its site. Adding a 'fees and costs' section (even as ranges: 'residential architect fees typically 8–15% of build cost') with a free feasibility review CTA would give AI engines citable, factual content for cost queries. This is the fastest route to citation on prompts 9-10.

Vendor Evaluation
3 prompts · 12 model responses analysed
Positive

GOAStudio appears in 2 of 3 vendor evaluation prompts across all 5 engines — exclusively driven by branded comparison (P2: GOAStudio vs Shape vs Scenario, 5/5 engines mention the brand) and conservation area RIBA query (P3: Gemini, Claude, and AIO mention GOAStudio; ChatGPT and Perplexity do not). On the unprompted 'Best residential architect London 2026' query (P1), GOAStudio appears in 0/5 engines — the brand does not break into AI-recommended shortlists when its name is not in the prompt. Scenario Architecture and Shape Architecture are cited by all 5 engines on P2, confirming their stronger ambient AI presence. The key win: when buyers know to ask about GOAStudio by name, all engines can find and describe it accurately.

Sentiment leaderboard

Share of voice across 10 prompts × 4 models
PosNeuAbs
1.
GOA Studioyou
2 · 0 · 8
2.
Scenario Architecture
2 · 0 · 8
3.
Urbanist Architecture
2 · 0 · 8
4.
Extension Architecture
2 · 0 · 8
5.
Shape Architecture
1 · 0 · 9

Frequently asked

What is a GAIO Deficit Report?

GAIO stands for Generative AI Optimization — getting your brand cited inside AI answers, not just ranked on a results page. The Deficit Report is RankBee's diagnostic: across leading AI engines (ChatGPT, Gemini, Perplexity, Claude, and Google AI Overviews) and a tailored prompt set, it shows which answers your brand is missing from, which competitors take the citation in your place, and the technical and content reasons why.

Who is this for?

Anyone whose audience now turns to ChatGPT, Gemini, Perplexity or Claude before making a decision. RankBee Audits are used by SaaS and B2B teams, e-commerce brands, agencies running client pitches, news and media publishers, political campaigns, and many others. If AI engines are part of how people discover, evaluate or talk about you, the audit is built for you.

How is this different from a traditional SEO audit?

A traditional audit grades you on Google's signals — backlinks, keywords, Core Web Vitals. RankBee grades you on what large language models actually reason about: entities, attributes, answer-first structure, citation-worthiness, and crawlability through the bot stack AI assistants use today (GPTBot, ClaudeBot, PerplexityBot, Google-Extended and 20 more). Strong Google rankings don't automatically translate into AI citations, and that gap is what the audit measures.

How does the audit work?

Four sections, each grounded in real data. Crawlability runs five technical phases: robots.txt rules, virtual-user probes from your target geographies, live LLM web-search fetches, bot-impersonation against your CDN, and token-depth indexability. Rankings Matrix runs your buyer prompts against up to 5 AI engines and logs every citation, co-citation, and competitor mention. Content Scorecard simulates AI ranking at the page level — RankBee ingests competitor content, generates variations, and scores yours 1–10 on the attributes models actually reward. Sentiment Snapshot reads how engines describe you when they do mention you, clustered by audience intent.

Where do the prompts come from?

RankBee discovers them for you. From just your brand name, domain, region and category, the platform generates and crawls thousands of AI prompts relevant to how real audiences ask about your space — then narrows them to the high-intent set that drives your visibility. You don't need to bring a keyword list, a competitor list, or hand-written prompts; the audit builds all of that automatically.

What does "invisible to AI" actually mean?

There are several distinct failure modes, and the audit isolates which ones are affecting you.

  • Uncrawlable. Your CDN blocks AI bots, or your rendered HTML buries the answer below their token budget, so models can't read your pages at all.
  • Crawlable but uncited. Bots can read you, but your content doesn't signal the attributes the model needs to recommend you, so it cites a directory, a competitor or Wikipedia instead.
  • Cited but mis-framed. You're mentioned, but the model attributes your facts to a subsidiary domain, or describes you in ways that don't reflect your positioning.
  • Locked out of live retrieval. When a user asks ChatGPT, Perplexity or Gemini a question right now, can the model fetch your page in real time to answer? The crawlability audit tests this end-to-end — many sites pass robots.txt but fail at the CDN or render layer, so live retrieval silently fails.
  • Excluded from training data. Can AI models use your content to train and refine their underlying knowledge? Your robots.txt and bot policies decide whether crawlers like GPTBot, ClaudeBot, Google-Extended and CCBot are allowed to ingest you. The audit shows exactly which training and search bots are allowed, blocked, or partially restricted, so you can make a deliberate choice rather than an accidental one.
How long does it take, and what do I need to provide?

Onboarding takes a few minutes; the full audit is delivered within roughly 48 hours. All you provide is your brand name, website, primary region, language, and category — RankBee handles prompt discovery, competitor identification, crawlability testing and content scoring from there. Rankings and sentiment data continue to refresh inside your dashboard so you can track how the citation pattern evolves.

What happens after the report — does it fix the issues?

The audit diagnoses; remediation happens in the rest of the platform. Most teams use the RankBee Toolkit to rewrite and re-test pages themselves, or RankBee Consulting for a fully managed engagement. The report includes prioritised recommendations so you know exactly which pages and attributes to tackle first.

Can I share the report with my team and stakeholders?

Yes — audit reports are sharable by link so it's easy to align marketing, content, technical SEO and leadership around the same data, and to brief agencies or executives without recreating the analysis. Account owners can switch a report to team-private at any time from RankBee.

How do I get a full audit?
Full audits are available to RankBee subscribers. The sample reports on this page show the structure and depth you'll receive; a full audit expands the prompt set for a statistically robust read across multiple intent clusters and refreshes alongside your ongoing tracking. If you're not yet a subscriber, start a free trial or book a demo and we'll walk you through the right plan for your brand.
Want this for your brand?

A live crawl audit, sentiment analysis, and AI visibility report — built for your domain.

This report runs a focused prompt set to show you the shape of the problem. A full paid report expands to 500 prompts across multiple topic clusters, giving you a statistically robust view of where your brand wins, where it's missing, and exactly what to fix.

Prepared by RankBee·rankbee.ai·RB-2026-05-11-0001