Earn the click
We earn rankings in Google and Bing for the searches buyers make when they are ready to choose a supplier.
- Sound technical foundations
- Pages built for every commercial query
- Content and links that make those pages rank
SEO & GEO service
Search engines and AI assistants now answer many of the questions that used to send buyers to websites. Our SEO and GEO work makes sure your brand is the one they find, cite and recommend, and that the visits you do get are built to convert.
Select a node to see its links
If you are not in the answer, you are not on the shortlist. That is why we plan the two together.
We earn rankings in Google and Bing for the searches buyers make when they are ready to choose a supplier.
We make your brand, pages and proof the sources AI assistants use when they answer, before a buyer clicks anything.
Why SEO and GEO belong together
SEO alone
GEO alone
One team, both
Ranking in search and being quoted by AI are two halves of one job. Run them as separate projects and the same research, rewrites and reporting get paid for twice. Run them as one plan and every piece of work feeds the next.
Book a discovery callThe waste you avoid
01 · Research
Two keyword studies and two competitor reviews of one market.
02 · Rewrites
One team optimises a page, and the other undoes it.
03 · Pages
Similar pages fight for the same searches.
04 · Facts
Different prices and claims that AI tools repeat back.
05 · Reporting
Two dashboards claiming the same results.
06 · Tech
Two crawls and two audits of the same site.
What we know
These 20 are existing SEO ranking factors. We tested how strongly each one moves AI answers, so SEO and GEO tasks that work together can be planned as one.
01
48.6% self-citation. Static corpus, no live retrieval.
98.4% citation rate across 914 responses, the highest of all four models.
Live web search. Zero mentions in 30 buying-intent prompts without source presence.
81.8% citation rate. Index-derived, and link authority gates the source pool.
02
Training cutoff. Newly published content has no direct impact.
Early 2024 knowledge cutoff. Recency within the training window helps moderately.
Live retrieval. Stale pages are invisible, and freshness is the primary signal.
Month-to-month mention rates fluctuate. Freshness is inherited from the index.
03
2.87 mean mentions per response, 3.1 times higher than Perplexity and Google.
1.35 mean mentions, higher than Perplexity despite similar citation rates.
58.4% zero-mention rate. Mention volume alone does not drive live retrieval.
Entity recognition is partly driven by mention frequency. Secondary to link signals.
04
Brands with national media coverage appear 2 to 3 times more in ChatGPT responses.
Editorial training data is curated for quality, and high-profile outlets are weighted heavily.
Live news crawl. A press piece from this week outweighs a feature from last year.
Google News-indexed outlets are cited above brand-owned pages for category queries.
05
YouTube is a top-5 cited domain in training data. 40.83% of cited videos have under 1K views.
YouTube transcripts in training data count regardless of view count.
YouTube is among the top-5 cited domains. Video descriptions and transcripts are retrieved live.
YouTube is Google-owned and the top-cited domain across AI Overview responses.
06
Best disambiguation of all four models. Wikipedia anchors are used in training.
Strong on direct entity queries, such as who a brand is, backed by Wikipedia content.
Wikipedia is accessible live and helps disambiguation, but less than for Google.
The Knowledge Graph is the primary entity resolution step and the highest-ROI technical signal.
07
Bulk low-authority mentions can substitute for fewer high-authority citations.
A DA-80 publication outweighs ten DA-20 blogs. Quality outweighs volume.
Cites only high-authority review and business sites. Low-authority sources are invisible.
PageRank-derived authority gates access to the AI Overview source pool.
08
Not retrieved live. Review text is absorbed into the training corpus only.
Transactional queries correlate with review platform presence, peaking at 1.46 mean.
2 to 3 times more appearances for brands with Trustpilot, G2 or Clutch presence.
Trustpilot, Which? and Google Business Profile are indexed directly as source material.
09
Author credentials in training data improve recall quality, but are not a hard gate.
Training favours credentialed, expert-authored sources, so author signals filter curation.
Author signals help domain credibility, but source authority outweighs the individual author.
96% of AI Overview citations come from E-E-A-T sources. Behaves as a near-binary gate.
10
Flat intent curve, with a 0.61 spread. Breadth beats intent-specific content every time.
37.7% zero-mention rate, lower than Perplexity at 58.4% because of quality depth.
Query-matched retrieval. Each page must match each query variant individually.
The Helpful Content system rewards topical authority and expert-authored category pages.
11
Comparison queries score lowest: 2.43 mean against 3.04 for informational, down 20 percent.
Weakest intent type, at 1.07 mean and 50% zero-mention. Describes categories, not brands.
1.38 mean on comparison, 53% above recommendation content. The highest-ROI content type.
Lowest zero-mention rate on comparison queries at 41.7%, against 55 to 60% for other intents.
12
2.68 mean on transactional queries. Moderate, with no deliberate avoidance.
1.46 mean, the only intent above its overall average. Zero-mention drops to 30.8%.
Defaults to aggregators on buying intent, with 0.93 mean and no direct vendor endorsement.
0.69 mean and 60.8% zero-mention. Deliberately routes transactional queries to paid ads.
13
Adding statistics raised LLM visibility by 33 percent against control, per the Princeton GEO paper.
Authoritative citations improved LLM citation rate by 115 percent for low-ranked content.
Specific data points are directly quotable, and Perplexity cites and attributes them live.
Helpful Content and E-E-A-T reward original analysis. Statistics signal expert authorship.
14
Consistent markdown with headers, numbered lists and bold brands stays stable across sessions.
Analytical prose works best, and comparison frameworks such as X vs Y align best.
Format is driven by the retrieved source documents, so your content format is irrelevant.
FAQ schema and a direct-answer structure match the AI Overview surfacing pattern.
15
0.61 mention-rate spread across intent types. Low qualifier sensitivity.
Qualifiers such as UK-based or affordable significantly shift the competitive landscape.
Qualifiers directly filter which live documents are retrieved for each query.
A Google Business Profile category can substitute for exact vocabulary in content.
16
0.61 spread across five intent types. Invest in breadth, not intent segmentation.
A 37 percent gap between transactional (1.46) and comparison (1.07). Intent matching pays off.
Comparison content is 53 percent above recommendation content, the clearest intent ROI of any model.
Comparison has 41.7% zero-mention against 60.8% for transactional, so comparison content wins.
17
Geographic qualification displaces mid-market brands, regardless of local presence.
Regional trade press and local directories modestly improve geo-qualified responses.
Live search responds to geographic footprint in near real time, so regional pages help directly.
Google Business Profile with LocalBusiness schema is the strongest geographic signal.
18
JSON-LD is stripped during training, so its value is indirect through Google indexation.
FAQ schema creates Q and A pairs that may be retained in response patterns.
Authority and freshness outweigh markup. Schema helps, but is not a primary signal.
Organization, FAQ and Review schema feed the Knowledge Graph and AI Overviews directly.
19
GPTBot for training and OAI-SearchBot for retrieval. Blocking either has separate consequences.
ClaudeBot for training and Claude-SearchBot for retrieval, the same dual-bot split as OpenAI.
PerplexityBot blocked means zero Perplexity citations. It is the only live retrieval path.
Blocking Googlebot removes the brand from the entire AI Overview source pool.
20
Training uses cached snapshots, so JavaScript-heavy sites are partly captured with loss.
Training data captures vary, and JavaScript-rendered sections are often absent.
Live fetch without JavaScript execution. JavaScript-only content is effectively invisible.
Google renders JavaScript, but indexing can lag by days to weeks for dynamic content.
What we do
An SEO and GEO agency that plans search and AI visibility together, so each piece of work supports the others.
01
We map the searches and AI questions buyers use at each stage, then score each one by commercial value.
02
Crawlability, site structure, indexation, speed, and entity and schema markup, so search engines and AI tools can read, connect and trust your pages.
03
The pages that should rank for each high-value query, with the proof, comparison and next step a buyer needs.
04
Content written to answer the questions buyers ask, structured so search and AI can lift the answer cleanly.
05
We test your brand across ChatGPT, Perplexity and Google AI Overviews with a fixed set of prompts, then close the gaps in sources, proof and comparisons.
06
We earn mentions on the publications and directories that both search engines and AI assistants rely on.
Beyond search
Search and AI data shows what buyers want. Every other channel works better when it uses it.
Book a discovery callOrganic rankings show where paid is buying clicks you already win. We move that budget to the gaps.
Search and AI questions show what buyers ask now, so content is planned from demand and written to be quoted.
Organic visitors arrive with clear intent, which gives conversion work a better base.
Search demand shows which problems buyers research, so paid social can target those themes.
The questions buyers ask in search and AI are the objections your email sequences need to answer.
We show which publications and communities shape your category, so PR earns the mentions that build visibility.
Links and citations
AI assistants name the brands they can cite. Our testing and published research point the same way: without third-party sources, a brand is rarely named. So we earn links and mentions on the sources AI already trusts.
One placement, two wins
The evidence
Buying-intent prompts in Perplexity returned zero brand mentions where the brand had no third-party source presence. Our testing.
Visibility gain for a lower-ranked site that cited its sources. Princeton GEO study, KDD 2024.
More appearances in Perplexity for brands on Trustpilot, G2 or Clutch. Our testing.
Correlation between branded web mentions and AI Overview visibility, against 0.22 for backlinks. Ahrefs study of 75,000 brands. Correlation, not proof of cause.
How we earn them
We test the questions buyers ask ChatGPT, Claude, Perplexity and Google AI Overviews, and record who is named and cited.
We see which sites each assistant cites for your competitors, and how often.
We rank sources by how often they are cited, their authority and their relevance to your category.
Digital PR, expert quotes, original statistics and review-platform profiles on those exact sources.
Network shows sites commonly cited by AI assistants in published citation studies. The mix varies by assistant, sector and prompt.
How we work
SEO has rankings and traffic. AI platforms report nothing. We start with a baseline of your visibility in AI answers, then measure SEO and GEO side by side every month.
The baseline · six steps
Pick the 3 to 4 commercial categories that drive revenue.
For each one: core topics, use cases, comparisons, problems and buyer questions.
Prompts that represent each cluster, not every possible question.
Run them across ChatGPT, Claude, Perplexity and Google AI Overviews, scored against the factors that move each model.
Topics you own, topics competitors own, and topics nobody owns yet.
The gaps become the roadmap. We re-run the same prompts to show what moved.
| Topic cluster | ChatGPT | Claude | Perplexity | AI Overviews | Status |
|---|---|---|---|---|---|
| Core service | 72% | 64% | 58% | 61% | You own |
| Use cases | 48% | 52% | 31% | 44% | Contested |
| Comparisons | 12% | 18% | 9% | 22% | Competitor owns |
| Buyer problems | 35% | 29% | 41% | 38% | Contested |
| Pricing and cost | 0% | 4% | 2% | 0% | Nobody owns |
Share of AI voice · all platforms
Gaps become the roadmap: comparisons and pricing first.
Re-tested every monthWhat we report
SEO · Demand capture
On the commercial queries you chose to win
Organic visits to the pages that matter
Enquiries and conversions by page and topic
Organic revenue and pipeline influenced
GEO · Demand influence
Share of tested answers that name your brand
How often AI links to your pages or sources
How often AI recommends you, not just mentions you
Your mentions against named competitors
A topic visibility framework, not a prompt library. Prompts are just how we measure it.
Get your baseline reportSEO helps your pages rank in search results. GEO helps your brand, pages and proof get used when an AI assistant writes the answer. The work overlaps, so we plan the two together.
Yes. AI assistants draw on the open web, and the pages they lean on are usually clear, well structured and already trusted by search. SEO builds that base, and GEO makes sure it is used.
No agency can promise a specific answer, because assistants change and can give different answers to different people. We make your brand easier to find, check and cite, then test the result with the same prompts every week.
We run a set of prompts in each assistant and record which answers name your brand. Visibility is the share of tested answers that name you, and we re-test every week. The free AI report shows the method.
It depends on your category, your starting point and how many sources need work. We set expectations after the first report, then review progress every week against the prompts we track.
Brands named in AI answers reach buyers earlier, while they are shortlisting. The same work usually strengthens organic rankings and gives paid and sales teams sharper proof. We report mentions and referrals next to revenue, so the impact is visible.
Use the checklist above. Ask which prompts they test, which measures they report and how they attribute results. Be cautious of guaranteed placements or claims that come without a method.
Yes, especially in niches where buyers ask specific questions. Smaller brands can win answers with well-targeted pages and specialist sources, even when they cannot match the coverage of larger competitors.
Direct answers, clear definitions, comparisons and sourced figures are usually the easiest for an assistant to quote accurately, so we build those first.
Organic data shows which keywords you already win, so paid focuses on the gaps. Paid data shows which terms convert, so SEO and content go after those next.
From the blog