Independent strategy example
An SEO and AI search strategy example for B2B SaaS
How I connect discovery, buyer evaluation, external evidence, technical priorities, and measurement—then turn the findings into a focused 90-day plan.
An independent exercise, not a client engagement. Nothing was implemented and no commercial outcome is claimed.
Research scope
Specific questions. Human-reviewed evidence.
- 24
- Buyer scenarios
- 4
- AI systems sampled
- 96
- Human-reviewed answers
- 44
- Evidence records reviewed
Search should support a buyer decision
Rankings and mentions matter, but the commercial job is to help the right buyer discover, evaluate, and choose the business.
Discover
Be present when a buyer first recognises the problem and explores the category.
Evaluate
Provide useful comparisons, proof, and answers for serious consideration.
Choose
Reduce uncertainty with credible evidence, clear product information, and usable conversion paths.
Discovery · Finding 1
A keyword count is not an audience.
One competitor ranked for 2,439 U.S. searches. About 74% of the traffic those rankings produced came from roughly five of them—login and brand searches from people who were already customers.
New-buyer demand
Deciding whether the product fits
Problem, category, comparison, integration, and proof questions. Searched by the people who sign off, not the people who log in.
Existing-user demand
Trying to use the product today
Login, setup, troubleshooting, and help questions. Genuinely useful, and a strong product signal—but a different audience.
Same brand, two different audiences—and only one of them buys. Measure them separately from day one.
Source context: third-party search positions and traffic estimates, U.S. sample. Limitation: a competitor estimate, not a measured figure, and not a claim about the company.
AI search · Finding 2
It gets into the conversation. It thins out at the shortlist.
Twenty-four buyer scenarios put through four AI assistants, 96 answers, every one read by a human.
| Stage | Count of 96 | Common track of 96 |
|---|---|---|
| Mentioned named in the answer | 36 | 37.5% |
| Considered seriously weighed up | 33 | 34.4% |
| Shortlisted makes the final few | 22 | 22.9% |
| Recommended picked as the answer | 21 | 21.9% |
The largest fall-off is not visibility—it is the step just before the shortlist.
Source context: controlled study, 24 designed buyer scenarios across four AI providers, 96 answers, all human-reviewed. Limitation: a designed sample on one date, not AI market share and not revenue.
AI search · Finding 3
Cited on the explaining, almost never on the choosing.
Google’s AI answer cites a page from this company on 329 searches—out of 5,031 unique ranking keywords, about 6.5% of everything it ranks for, worth roughly 76,700 monthly searches combined.
| Page type | Citations | Share of 329 |
|---|---|---|
| Blog and guides | 232 | 70.5% |
| Help centre | 31 | 9.4% |
| Homepage | 23 | 7.0% |
| Use cases | 11 | 3.3% |
| Download | 9 | 2.7% |
| Product and features | 6 | 1.8% |
| Other pages and subdomains | 17 | 5.2% |
The AI treats this company as a source of explanation, not as an option to choose.
Source context: AI citation export covering 5,031 unique ranking keywords. The page-type split sums to 329. Limitation: a citation is not a visit, competitors were not benchmarked on this, and the surface changes daily.
Competitive context · Finding 4
No one leads on every measure.
Two different questions, deliberately not averaged into a single league table.
Search presence
U.S. keywords each brand appears for
| Competitor A | 10,580 |
|---|---|
| This company | 7,231 |
| Competitor B | 6,537 |
| Competitor C | 2,439 |
AI visibility
Mentions in AI answers, 0–100
| Competitor A | 34 |
|---|---|
| Competitor B | 34 |
| Competitor C | 30 |
| Competitor D | 26 |
| This company | 23 |
Second of four on search presence, last of five on AI visibility. Those are not the same question.
Source context: third-party search and AI-visibility exports. Limitation: different pull dates and different instruments; neither figure is market share or buyer preference.
Authority · Finding 5
Fifteen industry publishers. None of them link here.
Filtered from ten thousand referring domains down to the publishers a buyer in this category would actually read.
- 83
- domains link to all four competitors and not to this company
- 15
- are genuine industry publishers after the directional gap is reviewed manually
- 5
- link to every one of the four competitors and not to this company
Ten more link to three of the four competitors. One of the fifteen alone reaches 251,000 subscribers in the buying function. The gap is named publishers, not a number of links.
Authority here is not the biggest score. It is the shortlist of publishers this buyer actually reads.
Source context: third-party backlink gap, followed by manual review of category-relevant publishers. Limitation: authority scores are directional vendor estimates, and a link is not coverage or a mention.
Content · Finding 6
It already ranks for the buying language. Just too deep.
Thirty-one buyer-relevant phrases where the company ranks beyond position 30 while at least one competitor sits inside the top 20.
| Buyer-relevant phrase | Estimated monthly searches |
|---|---|
| internal communication tools | 880 |
| employee experience platforms | 480 |
| internal communications platforms | 320 |
| employee engagement mobile app | 210 |
| frontline employee engagement | 210 |
It already ranks for all of them—this is a depth problem, not a coverage problem.
Source context: third-party keyword positions and estimated U.S. search volumes. Limitation: keyword-level, not topic-level; volumes are estimates and positions move.
Search distribution · Finding 7
Most rankings sit where buyers rarely look.
The site had 4,904 recorded rankings, but only 799—about 16%—sat on page one.
- 162
- Positions 1–3
- 637
- Positions 4–10
- 726
- Positions 11–20
- 1,834
- Positions 21–50
- 1,545
- Position 51+
Share of 4,904 recorded rankings
- 1–3 3.3%
- 4–10 13.0%
- 11–20 14.8%
- 21–50 37.4%
- 51+ 31.5%
Only 16% sat on page one. Around two thirds sat beyond page two.
Source context: third-party search reports. The three totals differ because they come from different reports and different days. 7,231 is the total U.S. keyword count from a competitor overview pulled on one date, and it drifts by a few dozen daily. 5,031 is the unique ranking keywords in the positions export. 4,904 of those carry an organic position, which is the set the distribution is measured against. Limitation: keyword counts are not unique visitors or demand, and positions move.
The strategy begins with decisions, not deliverables
The work was organised around what the business needed to understand and decide next.
- Where is the business visible today, and where does buyer demand go unanswered?
- Does existing-user support demand obscure new-buyer discovery?
- How is the company represented in AI answers and competitor shortlists?
- Which facts, sources, and outside signals support or weaken trust?
- Which content, technical, and authority priorities deserve the next 90 days?
- How should progress be measured without overstating causality?
Evidence Intelligence
Examine the information behind the answer
Evidence is one important layer of modern search, not a deterministic formula for controlling AI systems.
Facts
Are important product, company, security, integration, and customer claims clear?
Sources
Which owned and independent sources support those claims?
Corroboration
Do multiple credible sources agree, or does one unsupported statement carry the story?
Conflict
Where do outdated, ambiguous, or wrong-entity records create uncertainty?
Competitors
What credible proof exists for alternatives that the company lacks?
Authority
Which publishers, experts, directories, and communities shape category understanding?
Different stakeholders require different proof
One generic product page rarely answers every concern in a complex B2B purchase.
Operations
Adoption, frontline reach, workflow fit, and measurable operational value.
IT and security
Integration, controls, governance, deployment, and reliability.
People and communications
Engagement, accessibility, clarity, and content governance.
Finance and procurement
Business case, risk, implementation effort, and commercial proof.
The page that already works
The page can win attention and still lose the decision.
Three questions to ask of any page that already earns traffic, before building anything new.
Fit
“Does this fit an organisation like ours?”
Show sector and workforce relevance, and how it fits the existing stack.
Risk
“Can we roll this out confidently?”
Bring implementation, security, and integration detail within reach.
Proof
“Has it worked for someone like us?”
Put outcomes near the decision, and make them easy for a buyer and an AI answer to find.
Strengthen the page that already works before building new ones. Say what is good about it first, and mean it.
Limitation: this is a page-level assessment of the page already earning the most qualified attention. It is not a measured conversion result.
Turn the findings into a focused plan
The point of the analysis is to determine what should change, who needs to be involved, and what gets measured.
- Build buyer pages around real problems, category questions, comparisons, integrations, and proof.
- Strengthen company and product facts across owned pages and credible external sources.
- Close technical and internal-linking gaps that limit discovery or create poor journeys.
- Create a repeatable authority and digital PR programme around useful, defensible information.
- Align content, product, engineering, analytics, and communications around a shared 90-day roadmap.
Days 1–30
Establish the baseline
- Confirm priority buyers and questions.
- Fix critical technical and entity issues.
- Define measurement and ownership.
Days 31–60
Build the decision layer
- Create or improve priority buyer pages.
- Strengthen proof and internal linking.
- Begin outside-source and authority work.
Days 61–90
Test, publish, and learn
- Release the highest-value work.
- Repeat search and AI-answer sampling.
- Review movement and set the next roadmap.
Downloadable strategy example
Review the complete anonymised example
The 26-page PDF includes the research findings, study design, source limitations, technical and AI-study methods, priorities, 90-day roadmap, measurement, and guardrails.
What this demonstrates: how I move from specific findings to a plan while keeping estimates, observations, inferences, and unimplemented recommendations clearly separate.
Measurement
The guardrails behind every number reported
The reporting model distinguishes discoverability, engagement, progression, and business outcomes. It does not turn them into one convenient score.
- Visibility is not demand
- Ranking for something is not the same as anyone wanting it, and it is not intent to buy.
- A citation is not a visit
- Being named in an AI answer means the company was considered relevant. It does not mean anyone arrived.
- Engagement is not progression
- Reading the proof page is a good sign. It is not a commitment, and it will not be reported as one.
- Influence is not attribution
- Where organic contributed to a deal can be shown. A straight line from an AI mention to revenue cannot.
Signals are reported separately and never combined into a single score.
Measured
Captured directly.
Repeated pattern
Several sources point the same way.
Strong inference
It survives contradiction checks.
Hypothesis
Needs first-party validation.
- 44
- records reviewed
- 36 / 8
- independent versus owned, kept apart
- 10
- wrong-entity records quarantined
- 0
- composite scores calculated
Method appendix
How the findings were built and kept honest
The method matters because a strategy is only as useful as the definitions, exclusions, and checks behind it.
How technical findings earn priority
Fix what blocks a buyer, not what a crawler counts. The cycle is find, verify, fix, and re-check. A technical issue becomes a priority when it is affecting an important journey, reproduces, is material, has a clear owner, and can be validated after release.
Crawl occurrences are not affected users, lost conversions, or proof of a current defect.
How AI visibility was measured
24 scenarios, four providers, 96 answers, all human-read. The design varied one buyer concern at a time, kept question order stable, and recorded mention, consideration, shortlist, and recommendation separately.
| Stage | Observed |
|---|---|
| Mentioned | 36 |
| Considered | 33 |
| Shortlisted | 22 |
| Recommended | 21 |
One accepted run per pair, a sampled study, and no population or causal inference.
How confidence changes the language
Four confidence levels, and they are never blended. Measured evidence can use direct verbs. Repeated patterns need several sources. Strong inference must survive contradiction checks. A hypothesis stays a hypothesis until first-party data validates it.
The label decides the verb. A hypothesis does not get written up as a fact because it would read better.
Have a search or AI-discovery problem that needs this level of thinking?
Share the context, what matters to the business, and where the work is getting stuck.