Marketing Signals

Marketing Signals We Make Sure Your Brand Gets Found - Through Search & AI.

The old SEO measurement model has three layers: rank → traffic → conversions. It ran uncontested for twenty years.It's n...
01/06/2026

The old SEO measurement model has three layers: rank → traffic → conversions. It ran uncontested for twenty years.
It's now incomplete in ways that distort strategy.

The replacement has four layers:

𝗟𝗮𝘆𝗲𝗿 𝟭 - 𝗩𝗶𝘀𝗶𝗯𝗶𝗹𝗶𝘁𝘆 (𝘄𝗵𝗲𝗿𝗲 𝘆𝗼𝘂 𝗮𝗽𝗽𝗲𝗮𝗿) Not traffic - impressions. Do buyers encounter your brand at all, on Google and in AI engines? Measured by impression share by query type plus Share of AI Voice across ChatGPT, Claude, Perplexity, Gemini.
𝗟𝗮𝘆𝗲𝗿 𝟮 - 𝗖𝗶𝘁𝗮𝘁𝗶𝗼𝗻 (𝗵𝗼𝘄 𝘆𝗼𝘂'𝗿𝗲 𝗰𝗿𝗲𝗱𝗶𝘁𝗲𝗱) When AI engines answer, do they credit your URL? Citation builds authority and brand recall even when buyers don't click. Measured by citation rate and citation share vs competitors.
𝗟𝗮𝘆𝗲𝗿 𝟯 - 𝗜𝗻𝗳𝗹𝘂𝗲𝗻𝗰𝗲 (𝗱𝗼𝘄𝗻𝘀𝘁𝗿𝗲𝗮𝗺 𝗯𝗿𝗮𝗻𝗱 𝗲𝗳𝗳𝗲𝗰𝘁) What happens after buyers see your brand cited, without a click? Measured by branded search trend, direct traffic trend, and AI-source self-reporting on demo calls.
𝗟𝗮𝘆𝗲𝗿 𝟰 - 𝗖𝗼𝗻𝘃𝗲𝗿𝘀𝗶𝗼𝗻 (𝘁𝗵𝗲 𝗳𝘂𝗻𝗻𝗲𝗹 𝘀𝘁𝗶𝗹𝗹 𝗲𝘅𝗶𝘀𝘁𝘀) Clicks are concentrating into higher-intent queries. AI-introduced visitors convert at different rates to search-introduced visitors. Measured by conversion rate by source - segmented properly, not blended.

The strategic value: Layer 4 (conversion changes) is caused by decisions made 60–90 days upstream in Layers 2 and 3. Teams monitoring only conversion are always reacting. Teams monitoring all four layers can act before the downstream effect arrives.

The full measurement framework with specific metrics, tools, and cadence for each layer:

The B2B buyer journey isn't a funnel any more.It's a series of AI-mediated micro-decisions across ChatGPT, LinkedIn, Red...
30/05/2026

The B2B buyer journey isn't a funnel any more.

It's a series of AI-mediated micro-decisions across ChatGPT, LinkedIn, Reddit, podcasts, G2 and analyst notes — happening before any tracked touchpoint fires in your attribution model.

By the time someone fills your demo form, the shortlist was set weeks ago by sources you can't see.

Part 4 of our 12-part Visibility Playbook Series.

This week: B2B lead generation.

The playbook covers:

→ Why "How did you hear about us" is now the most honest attribution signal you have
→ Solution and comparison pages built for both ranking and citation
→ The third-party citation surfaces compounding for B2B in 2026 — and the ones quietly losing influence
→ A measurement framework that works on a 6–18 month sales cycle

3,700 words. Written for marketing leaders who can no longer pretend last-click attribution tells the truth.

→ Link to the full playbook in the first comment ↓

If I could install one thinking framework for someone and walk away, it would be Charlie Munger's.His "mental models + i...
29/05/2026

If I could install one thinking framework for someone and walk away, it would be Charlie Munger's.

His "mental models + inversion" approach catches the widest range of thinking failures across the widest range of contexts. Investing, strategy, creative decisions, leadership, writing - every domain benefits from:

→ "How would this fail?" (inversion)
→ "What am I missing because of the way I've framed this?" (mental models)
→ "What would I need to believe for this to be wrong?" (systematic doubt)

These are generalisable. Munger's framework isn't just for value investing - it's a discipline for avoiding the kind of reasoning mistakes that smart people make specifically because they're smart enough to rationalise anything.

As a Claude Skill: load it once, and Claude auto-invokes it whenever you're asking about decisions, evaluations, strategies, or things you're about to commit to. You don't have to remember to ask "but what could go wrong?" Claude asks it for you.

£4.99 at authority(.)md. Installs in 60 seconds. Works across Claude(.)ai, Claude Code, and the API.
What a Claude Skill actually is and how to install one:

There's a measurement problem at the centre of most GEO audits that nobody talks about publicly.It's called the cold-sta...
28/05/2026

There's a measurement problem at the centre of most GEO audits that nobody talks about publicly.
It's called the cold-start problem.

Run an experiment: open a new ChatGPT chat and ask "What's the best CRM for a 50-person SaaS team?" Note the answer. Then in the same conversation, ask "What about for nonprofits?" The second answer is biased by the first. The model has already built context. It assumes continuity.

This means: every prompt you run after the first one in a session is contaminated. You can only get a clean, unbiased measurement of how a brand surfaces in AI answers if every prompt is asked in a completely fresh session - zero prior context, no carried history, no priming.

Most tracking tools batch multiple prompts into one session because it's cheaper and faster. The data comes back biased in ways you can't see or correct for after the fact.

What clean methodology looks like: one prompt, one fresh session, session closed, next prompt opened in a brand-new session. Slower. More expensive. The only way to get an uncontaminated reading.

Before you accept any AI visibility report, ask one question: "How were sessions managed?" If the answer isn't cold-start per prompt, the data has a structural problem that no amount of post-processing fixes.

Seven sources of measurement noise in LLM tracking, and how each one is controlled for:

GEO

Five mistakes that kill the four-day-week model. Each one is avoidable. Each one has killed at least one experiment I've...
28/05/2026

Five mistakes that kill the four-day-week model. Each one is avoidable. Each one has killed at least one experiment I've watched.

Treating it as a perk. Underpricing the operational redesign. Weak leadership discipline. Sloppy client communication. Using it as a cost-cutting move.

Get all five right and the model works. Trip on any one and it fails inside six months.

Read it → https://garethhoyle.com/writing/why-most-four-day-experiments-fail

Semrush says your authority is 49. Ahrefs says 53. Moz says 44. DataForSEO says 51.Which one is right?All of them, sort ...
27/05/2026

Semrush says your authority is 49. Ahrefs says 53. Moz says 44. DataForSEO says 51.Which one is right?
All of them, sort of. None of them, exactly.

Each metric measures roughly the same underlying thing - link-based authority - using different crawls, different link-quality models, and different scoring functions. They diverge predictably, not randomly. And understanding how they diverge tells you something useful.

The consistent offsets to know:
→ 𝗔𝗵𝗿𝗲𝗳𝘀 𝗗𝗥 runs 3–5 points higher than 𝗦𝗲𝗺𝗿𝘂𝘀𝗵 𝗔𝗦 for established domains (broader crawl, pure link focus, no organic-traffic adjustment)
→ 𝗠𝗼𝘇 𝗗𝗔 runs 3–7 points lower (stricter spam filtering, ranking-prediction calibration)
→ 𝗗𝗮𝘁𝗮𝗙𝗼𝗿𝗦𝗘𝗢 𝗗𝗼𝗺𝗮𝗶𝗻 𝗦𝗰𝗼𝗿𝗲 tracks Ahrefs closely within ±2 points typically

The rule that comes from this: when all four agree within ~5 points, the consensus is robust. Act on it.

When variance exceeds 8 points, dig deeper - one of seven specific patterns is almost always the explanation, and each pattern changes which tool to trust.

The CFO-defensible answer is a consensus, not one tool's view. The analysis that surfaces what single-source measurement misses:

A confession from someone who sells thinking frameworks: The worst outcome isn't buying the wrong framework.It's spendin...
27/05/2026

A confession from someone who sells thinking frameworks: The worst outcome isn't buying the wrong framework.
It's spending three weeks debating which to buy and never installing anything.

The frameworks are £4.99. The downside risk of the "wrong" one is £4.99 and ten minutes. The cost of continued default-mode Claude - no framework, no consistent mental model applied, same output you've always had - is harder to quantify but almost certainly larger.

Practical advice for anyone stuck in the selection loop:

Start with Munger. If you genuinely can't decide, Munger's mental models and inversion catches the widest range of thinking failures across the widest range of contexts.

You cannot go wrong as a starting point. From there, the next buy is informed by using the first one - you'll know after a week what kind of challenge or complement you want next.

Buy one. Install it. Use it on a real decision. The shift in how Claude reasons is immediately detectable.

If it's useful, buy a second. If it isn't, £4.99 is cheap tuition in what doesn't work for you.

The honest buyer's guide:

Of the eight reasons your competitor is winning the AI conversation, seven take 3–18 months to fix.One takes a week.Buil...
26/05/2026

Of the eight reasons your competitor is winning the AI conversation, seven take 3–18 months to fix.
One takes a week.
Build the comparison pages.

When AI engines synthesise "X vs Y" or "best X for Y" answers, they retrieve comparison content, read it, and use that framing in their answer. Whoever published the most authoritative comparison page owns the framing.

If your competitor has a "them vs you" page and you don't, the AI uses their framing. Every. Single. Time.
What to build:
→ /vs/[competitor] pages for each major competitor
→ An /alternatives/ page covering your category

Structure:
one clear claim per section, facts first, fair trade-off acknowledgements (AI engines cite fair comparisons over pure marketing)

Publish them. AI engines can start citing your framing within weeks.

Compare that to the slow fixes:
→ Wikipedia presence: 6–12 months minimum to earn notability
→ Editorial volume gap: years to close if they have a structural lead
→ Reddit reputation: multi-year, can't be rushed
→ Sentiment frame: 6–18 months to shift

The comparison page is the only high-leverage GEO move in existence that can be shipped on a Monday and start working before the end of the month.

Full eight-reason diagnostic at the link.

A full case study of one specific agentic system we shipped. amivisible(.)co. AI-search-visibility audits.Two people, tw...
26/05/2026

A full case study of one specific agentic system we shipped. amivisible(.)co. AI-search-visibility audits.

Two people, two weeks of work. Now it's an agent pipeline plus a human review. Same output. Different economics.

The post walks through architecture, failure modes, and what we'd build differently next time. If you want to see what 'agentic' actually looks like in production, this is it.

Read it → https://garethhoyle(.)com/writing/anatomy-of-real-agentic-workflow

Two situations where Custom GPTs are the better choice over Claude Skills, even if you use both platforms:1. Your team i...
25/05/2026

Two situations where Custom GPTs are the better choice over Claude Skills, even if you use both platforms:

1. Your team is locked in the ChatGPT ecosystem. If 10 people on your team use ChatGPT and one uses Claude, the Custom GPT is the right format - it's shareable across the team in one click. The Claude Skill is the better format, but only reaches the one person.

2. You need Actions - external API calls from the assistant. Custom GPTs can call external APIs directly from the chat interface. Want an AI assistant that can read from your CRM, query a database, or call a webhook while applying a Buffett framework to the output? Custom GPT with Actions is the path. Claude Skills shape how the AI thinks; they don't give it the ability to do things. That's MCP for Claude, Actions for ChatGPT.

The limitation that bites people using Custom GPTs for frameworks: the Instructions field caps at ~8,000 characters.

Most thinking frameworks fit with room to spare, but if you're trying to load multiple frameworks or a very detailed one, you'll be pushed into Knowledge files - which are slower and less "always on."

The full honest comparison - including when system prompts beat both:

Address

Manchester

Opening Hours

Monday 9am - 5pm
Tuesday 9am - 5pm
Wednesday 9am - 5pm
Thursday 9am - 5pm
Friday 9am - 5pm

Telephone

+441619288799

Website

Alerts

Be the first to know and let us send you an email when Marketing Signals posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Contact The Business

Send a message to Marketing Signals:

Shortcuts

Share

Category