GEO Questions, Answered One at a Time on MarGen | AI Visibility Agency

Do you guarantee results?

Sat, 29 Aug 2026 23:30:00 +0000

The short version

No agency controls model outputs, so no agency can guarantee a citation.
A guarantee is usually either meaningless in the small print or a compliance risk.
What can be committed to: scope, method, measurement, reporting and responsiveness.
In regulated sectors an unsubstantiated performance guarantee is a problem in its own right.
The right question is not “do you guarantee” but “what happens if it does not work”.

Why the guarantee cannot be real

Citation depends on the retrieval and selection behaviour of systems operated by OpenAI, Google, Anthropic, Perplexity and Microsoft. Those systems change without notice, they weight sources differently from each other, and they produce different answers to identical questions on different days.

How does GEO compare to PPC spend?

Sat, 29 Aug 2026 23:00:00 +0000

The short version

PPC is rented. Turn it off and visibility ends the same day.
GEO compounds. Work done in month one is still working in month eighteen.
GEO is slower to start. Expect three to six months before meaningful movement.
Cost per lead usually crosses in GEO’s favour somewhere between six and twelve months.
PPC is still correct for launches, tests, seasonal pushes and immediate pipeline.

The real difference

The distinction is not channel against channel, it is expenditure against investment.

What about podcasts?

Sat, 29 Aug 2026 22:00:00 +0000

The short version

Audio is not read. Transcripts and show notes are.
Guest appearances usually beat hosting your own show, at least at first.
A guest slot is third-party corroboration on somebody else’s domain, which is what models weight.
Episode pages with full transcripts become genuinely citable assets.
Hosting your own show is a long, expensive route to authority if nobody is listening yet.

Guest appearances versus your own show

This is the distinction that matters commercially, and it usually points the opposite way to instinct.

Does video content help GEO?

Sat, 29 Aug 2026 21:00:00 +0000

The short version

The text around the video does the citation work, not the footage.
A full transcript on the page is the single highest-return change.
Auto-generated captions are a weak substitute: unpunctuated, often wrong on technical terms.
YouTube has additional weight in Google’s AI surfaces because of the ownership relationship.
Video is strong for trust and conversion, which is a separate and legitimate reason to make it.

Why the transcript is the asset

When an assistant answers a question, it is working with text. If your expertise exists only as speech inside a video file, it is not in a form the retrieval step can use, however good it is.

How often do I need to publish?

Sat, 29 Aug 2026 20:00:00 +0000

The short version

Sustained cadence beats total volume. Stopping undoes more than slowing down would have.
Roughly weekly is a good target for most firms. Two a month is a workable floor.
Restructuring existing pages counts, and usually returns more per hour than new writing.
Output beyond your capacity to keep it specific and accurate is counterproductive.
Named-author output is worth more than the same volume published anonymously.

Why consistency beats bursts

Citation authority accumulates through repeated corroboration over time. A model does not decide you are authoritative because thirty pages appeared in one week. It builds confidence as material accumulates, gets referenced, and holds up.

Does social media affect GEO?

Sat, 29 Aug 2026 19:00:00 +0000

The short version

Social posts themselves are rarely cited directly in generated answers.
Profiles do real entity work: they corroborate who you are and what you do.
Distribution matters. Reaching the people who write citable sources is the indirect route.
Forums and community platforms are the genuine exception and are cited heavily.
LinkedIn carries disproportionate weight in B2B because of named-person authority.

Where the effect actually comes from

Three distinct mechanisms get muddled together under “does social help”, and they behave very differently.

Do I need to be on Wikipedia?

Sat, 29 Aug 2026 18:00:00 +0000

The short version

Wikipedia is unusually influential in entity resolution, so an article genuinely helps.
Most businesses do not meet the notability guidelines, and that is not a failure.
Creating an article about yourself is against the guidelines and tends to backfire publicly.
Wikidata is a separate, more achievable project and delivers much of the entity benefit.
Consistent facts across many ordinary sources substitute for it well.

Why Wikipedia carries weight

Wikipedia is disproportionately represented in the material language models were trained on, and it is structured, cross-referenced and heavily edited. For resolving what an entity is and what it is associated with, it is close to ideal, so models lean on it.

Does my existing content help?

Sat, 29 Aug 2026 17:00:00 +0000

The short version

Existing content is usually an asset that is badly packaged, not a liability.
Restructuring beats rewriting on cost, speed and risk.
The common failure is format, not substance: the answer is present but buried.
Some content genuinely cannot be saved, and knowing which is the point of an audit.
Thin, generic or undated pages are the ones to retire rather than rework.

Why restructuring usually wins

The typical professional services site has years of accumulated material written by people who genuinely know the subject. It was written for a reader scrolling, and often for a keyword. It opens with context, builds an argument, and arrives at the useful bit somewhere in the middle.

Can you A/B test GEO?

Sat, 29 Aug 2026 16:00:00 +0000

The short version

No split testing. You cannot serve two versions of a page to one model and compare.
Model outputs vary between identical runs, so a single comparison proves nothing.
Before-and-after on a fixed query set does work, with enough runs to see through the noise.
Cohort testing across comparable pages is the closest thing to a real experiment.
Always keep an unchanged control group. Platforms shift on their own and will fool you.

Why the conventional method breaks

A/B testing needs three things: two variants live at once, traffic randomly split between them, and a stable measurement of what each produced. AI citation gives you none of them.

What is a good AI citation rate?

Sat, 29 Aug 2026 15:00:00 +0000

The short version

20 to 30 per cent of relevant queries is strong in most professional services.
50 per cent or more is achievable in narrow, less contested niches.
The rate is meaningless without the query set it was measured against.
Which queries you are cited for matters far more than how many.
Your only genuinely useful benchmark is your own baseline plus your named competitors.

Why the headline number misleads

A citation rate is a fraction, and almost all the argument is in the denominator. Cited in 60 per cent of queries sounds excellent until you learn the query set was twelve terms your firm already dominated. Cited in 15 per cent sounds weak until you learn it covers 200 competitive buying-intent queries across five platforms.

What is Retrieval-Augmented Generation?

Sat, 29 Aug 2026 14:00:00 +0000

The short version

A model without retrieval answers from training data: fixed, undated, uncitable.
RAG adds a fetch step, so the answer is grounded in real sources that can be named.
Two gates: your content must be retrieved, then it must be selected for the answer.
Citations only exist because of retrieval. No retrieval, no link, no attribution.
Optimising for retrieval and optimising for selection are different jobs.

How it works, without the jargon

A language model on its own is a fixed snapshot. It has read an enormous amount up to a point in time and can produce fluent text from that, but it cannot tell you what changed last week and cannot honestly attribute a claim to a source, because it is not reading anything as it answers.

Do AI models use the same signals as Google?

Sat, 29 Aug 2026 13:00:00 +0000

The short version

Substantial overlap: relevance, authority, crawlability, credible sources.
AI weights extractability much higher. If an answer cannot be lifted cleanly, it is not used.
AI weights factual specificity higher. Vague claims are hard to quote and rarely quoted.
Engagement metrics matter much less. There is no click to measure in a generated answer.
The practical consequence: ranking well is not evidence of being cited.

Where they agree

Most of the foundation is shared, and this is genuinely good news because it means existing SEO investment is not wasted. Both systems need to reach your page, both prefer content that matches what was asked, both use signals of authority and trust, and both are influenced by what credible independent sources say about you. Analyses consistently find that a large majority of AI-cited URLs also rank well conventionally. Strong SEO is close to a precondition rather than a competitor.

What is entity authority?

Sat, 29 Aug 2026 12:00:00 +0000

The short version

Two components: identification (who you are) and association (what you are known for).
Identification comes first. An unresolved entity cannot accumulate authority on anything.
Association is built mostly off your own site, through independent corroboration.
Consistency across sources matters more than volume on any one source.
It is the slowest-moving and most defensible part of AI visibility.

Identification and association

Identification is whether a model can confidently say your business exists and is distinct from everything else with a similar name. This is where most problems live and where almost nobody looks. A company trading under three variations of its own name, with a different description on its website, its LinkedIn and its Companies House filing, is not one clear entity to a model. It is a cluster of weak signals that may or may not refer to the same organisation.

Is GEO just a buzzword?

Sat, 29 Aug 2026 11:00:00 +0000

The short version

The scepticism is reasonable. Search marketing renames itself roughly every three years.
The behaviour change underneath it is real and measurable, not a projection.
Most of the work is recognisable: entity clarity, structured data, content quality, corroboration.
The genuinely new part is the outcome. There is no list, so there is no position ten.
Be sceptical of anyone selling GEO as entirely separate from SEO with its own separate retainer.

What is actually new, and what is not

Start with the honest part. A large majority of GEO work would be recognised by any competent SEO practitioner. Making an entity unambiguous, marking content up properly, writing so the answer comes first, earning credible third-party mentions: none of that was invented in the last two years, and anyone claiming otherwise is selling.

Is GEO only relevant for B2B?

Sat, 29 Aug 2026 10:00:00 +0000

The short version

B2B adopted first because B2B buying is research-heavy and committee-driven.
Consumer AI search is growing fast, particularly for considered and high-value purchases.
Consumer exposure concentrates in categories where people already researched before buying.
Low-consideration and impulse retail remains largely unaffected for now.
The B2B lead is a timing difference, not a permanent structural one.

Why B2B moved first

B2B buying has the properties AI assistants are good at. The purchase is expensive, the decision is made by several people, the buyer is often not an expert in what they are buying, and there is a genuine shortlisting stage before anyone speaks to a supplier. An assistant that returns three credible named options with reasons is genuinely useful in that situation.

Who needs GEO?

Sat, 29 Aug 2026 09:00:00 +0000

The short version

If your buyers research before they contact you, they are probably researching in an assistant.
Considered purchases need it most. Impulse and pure-local purchases need it least.
High-value, low-frequency, comparison-heavy buying is where AI shortlisting bites hardest.
The honest test takes two minutes and is at the bottom of this page.
Sector matters less than buying behaviour. Two firms in the same sector can have very different exposure.

Why some businesses need this and others do not

The useful question is not “is my industry affected”. It is how much research happens before a buyer contacts you, and where that research now takes place.

Why doesn't my business show up in ChatGPT?

Sat, 29 Aug 2026 08:00:00 +0000

The short version

OpenAI runs three separate crawlers. Blocking or missing one is not the same as the others.
ChatGPT answers either from training memory or from live browsing, and they fail differently.
If it browses and still misses you, the problem is extraction or corroboration, not access.
If it answers confidently and wrongly, that is a training-memory problem and it is slower to fix.
Bing presence remains a sensible hedge, because OpenAI’s retrieval has historically leaned on it.

Why this happens

The most common mistake is treating ChatGPT as one system. It is at least two, and they fail for different reasons.

Why doesn't my business show up in Perplexity?

Sat, 29 Aug 2026 07:00:00 +0000

The short version

Perplexity shows its sources, so you can see exactly who is being cited instead of you.
Two agents: PerplexityBot for indexing and Perplexity-User for user-triggered fetches.
It leans harder on recency than the other assistants. Stale pages lose here first.
Forums and community threads are cited heavily, often above company websites.
It is usually the fastest platform to show movement, which makes it a good early indicator.

Why this happens

Perplexity is the most transparent of the major assistants, and that is a gift for diagnosis. Every answer lists its sources. You are not guessing at why you were omitted, you can read what was used instead.

Why doesn't my business show up in Google AI Overviews?

Sat, 29 Aug 2026 06:00:00 +0000

The short version

Ranking in classic Google is close to a precondition. This is the biggest single difference.
Google-Extended does not control AI Overviews. It governs Gemini training and grounding.
You cannot opt out of AI Overviews separately from Google Search.
Overviews favour pages that answer a sub-question cleanly, not necessarily the top result.
Zero clicks with high impressions is often success here, not failure.

Why this happens

AI Overviews are built on Google’s existing index rather than a separate one. That has a blunt practical consequence: if you do not rank reasonably well for the query underneath the Overview, you will rarely be drawn into it. Analyses consistently find a large majority of cited URLs already ranking on page one.

Why doesn't my business show up in Gemini?

Sat, 29 Aug 2026 05:00:00 +0000

The short version

Gemini and AI Overviews are both Google, but they are separate surfaces with different answers.
Google-Extended does apply here, unlike AI Overviews. A block genuinely excludes you.
Grounding runs through Google Search, so classic visibility is still the foundation.
The Workspace version can also draw on a user’s own documents, which you cannot influence.
Entity clarity matters unusually much, because Google has a knowledge graph to check you against.

Why this happens

The first thing worth separating is that Gemini and AI Overviews are not the same product, even though both are Google and both generate answers. Overviews appear above search results. Gemini is a standalone assistant with its own interface, apps and Workspace integration. They can and do give different answers to the same question, so testing one tells you less about the other than people assume.

Why doesn't my business show up in Claude?

Sat, 29 Aug 2026 04:00:00 +0000

The short version

Claude is more conservative about naming businesses than ChatGPT or Perplexity.
Two agents: ClaudeBot for training data and Claude-User for user-triggered fetches.
It leans toward established, verifiable sources over recent or promotional ones.
In regulated categories it will often decline to recommend anyone at all.
Corroboration matters more here than volume of your own content.

Why this happens

Claude behaves differently enough from the other assistants that firms often assume something is broken when nothing is.

Why doesn't my business show up in Copilot?

Sat, 29 Aug 2026 03:00:00 +0000

The short version

Copilot grounds in Bing. Bing visibility is the lever, and it is usually the neglected one.
Bing Webmaster Tools is free, takes an afternoon, and most UK firms have never claimed it.
Bing and Google rank differently. Assuming Google performance transfers is a mistake.
Enterprise Copilot also grounds in the customer’s own Microsoft 365 data, which is closed to you.
Copilot reaches business users inside the tools they already work in, which raises the intent.

Why this happens

Copilot is the most straightforward of the six to diagnose, because the dependency is explicit. It is grounded in Microsoft’s search index. If Bing does not have you, or ranks you poorly for the query, Copilot will not surface you.