Wayfinder AI
Compass
All Comparisons

The AI & AEO Tool Stack: What We Use and Why

A practical stack for AI search work: frontier models for strategy, local models for batch execution, and specialised tools for AI navigation audits.

claudechatgptgeminilocal-modelscompassscreaming-frogprofound

Most "AI for SEO" content stops at "use ChatGPT to write meta descriptions." The real question is what tools you need across the full AI search workflow — strategy, analysis, content, code, and technical auditing — and where it is worth paying versus building yourself.

The short answer: you probably need fewer tools than you think. A solid stack for most teams is Screaming Frog (general crawlability), Compass (AI-specific navigation testing), your existing SEO platform (Ahrefs or Semrush), and one frontier model (Claude) for strategy and code. Everything else is optional.

AI Platforms for Strategy and Execution

Claude

The strongest all-rounder for complex reasoning, strategy, and production code. It maintains context across long conversations and pushes back when a premise does not make sense. We use Claude Code for the bulk of our development work.

  • Best for: site audits, content strategy, competitor analysis, production scripts
  • Limitation: cost scales with volume

ChatGPT / ChatGPT Search

Good for quick retrieval, validation, and second opinions. ChatGPT Search is useful for real-time web retrieval, though the analysis tends to be surface-level compared with Claude.

  • Best for: fast answers, API-driven workflows, web retrieval
  • Limitation: less depth for strategic reasoning

Gemini

Best-in-class for web search and image generation. Instruction-following can be inconsistent for multi-step analysis, but it is genuinely useful for research and visual work.

  • Best for: web retrieval, SERP analysis, image generation
  • Limitation: weaker for deep analysis and production code

Local models

A dedicated AI workstation (£3–4k upfront) running a model like Qwen 3.5 122B can handle batch execution, data processing, and template-based content generation with no per-token cost. The workflow that works for us: a frontier model writes detailed briefs; a local model executes them at scale.

  • Best for: batch work, sensitive data, high-volume execution
  • Limitation: requires hardware investment and technical maintenance

AEO Audit and Visibility Tools

Compass (Wayfinder)

Purpose-built for AI navigation testing. Compass simulates how AI agents actually navigate your site — testing entry strategies, measuring click depth, identifying loop traps, and mapping where agents succeed and fail. It answers the question Screaming Frog cannot: can an AI agent find this page by navigating your site?

  • Best for: AI-specific navigation audits, technical accessibility for agents
  • Pricing: from £50/month

Screaming Frog

The industry-standard crawlability tool. It tells you whether any bot can see your pages, compares raw HTML to rendered DOM, maps link structure, and surfaces JavaScript rendering issues. Everyone doing technical SEO should have it.

  • Best for: general crawlability, rendering checks, link structure
  • Pricing: free version available; paid licence £199/year

Prompt visibility trackers (Profound, Otterly, Peec AI, etc.)

These track how often your brand appears in AI responses. They look good and do interesting things, but we have argued extensively that prompt-position metrics are noisy — LLMs are non-deterministic and most tools do not sample enough to be statistically valid. Frequency over 60–100 runs per prompt is the only meaningful measure, and few tools do that.

  • Best for: brand mention tracking at scale
  • Our take: optional, not essential; a self-built API tracker gets you 80–90% of the insight

SEO platforms with AI features

Ahrefs and Semrush are adding AI visibility reporting. If you already use one, this is the lowest-friction place to add an AI layer. The features are not as deep as dedicated tools, but they cover the basics.

  • Best for: adding AI reporting to an existing SEO workflow
  • Our take: sensible if you already subscribe; do not buy a platform just for this

The Self-Build Option

If you have technical resource, you can build a lot of this yourself:

  • Prompt tracker: query the OpenAI/Anthropic/Gemini/Perplexity APIs, parse responses for brand mentions, run 60–100 times per prompt, store results.
  • Basic crawlability testing: Python libraries like Scrapy or LibreCrawl can compare raw HTML to rendered content.
  • Manual AI navigation testing: follow the DIY audit process — send AI agents to your site with specific tasks, record the results, score them.

What is harder to build: ML-based content scoring, large-scale automated navigation testing, and polished UIs for non-technical users. For prompt visibility, self-build makes strong economic sense. For technical accessibility, the existing tools exist because the problem is harder — but a basic version is still buildable.

Recommended Stacks by Team

Solo practitioner / small team

  • Claude Pro (£90/month)
  • Screaming Frog free + manual AI testing
  • Compass for AI navigation audits
  • Self-built prompt tracker if needed
  • Total: ~£140/month

Growing team / agency

  • Claude for strategy and complex work
  • API access for automated workflows
  • Screaming Frog + Compass
  • Optional prompt visibility tracker
  • Total: £200–500/month

Technical team willing to invest

  • Claude for frontier reasoning
  • Local 122B model for batch execution (£3–4k one-time)
  • Screaming Frog + Compass
  • Self-built tooling for tracking and workflows
  • Total: £3–4k upfront + ~£100/month ongoing

What to Prioritise

Start with technical accessibility: can AI find your content? Then content quality: is the content good? Prompt visibility is last — it tells you the least and costs the most. Once your foundations are sound, use the technical AEO checklist to close any gaps.