Meliada
Answer engine optimisation

Be the answer, not a result.

Your buyers stopped opening ten tabs. They ask once, read one answer, and shortlist whoever it names. We make that name yours.

Request an audit See how it works
AI answerIllustrative
› Which warehouse can I migrate to without downtime?
Three platforms support live cutover. The approach that avoids downtime uses dual-write with a verification window before the switch.

Sources cited

yourdomain.comCited
competitor-a.com
competitor-b.com
4 prompts won 3 gaps open

The mechanism

Answer engines do not rank pages. They pull passages, weigh which one settles the question, and name a couple of sources. Every stage rewards writing that commits.

01

A question is asked

Longer than a search query, and it usually carries the buying context already.

02

Passages are pulled

Not whole pages. Whatever answers cleanly on its own gets retrieved.

03

Sources are weighed

Evidence inside the passage, and corroboration across the open web.

04

Two get named

They take the citation and the click. Everything else was retrieved for nothing.

The gap

Coverage clusters where you already publish and vanishes exactly where the decision gets made. The audit shows you which side of zero you sit on.

Specimen audit Citation share against the leading competitor Diverging bar chart across seven topic clusters. Bars to the right in green show clusters where the brand is ahead of the leading competitor; bars to the left in red show clusters where it is behind. Implementation leads by 24 points; benchmarks trails by 34 points. Citation share against the leading competitor Implementation Integrations Pricing and cost Comparisons Security and compliance Migration Benchmarks +24 -34 -40 -20 0 +20 +40 Percentage points vs the leading competitor
Ahead of the leader Behind the leader
Illustrative figures. This is the shape of the audit, not a result we are promising. Your own numbers land in week three.
View as table
Citation share vs the leading competitor
Topic clusterPositionPoints
ImplementationAhead+24
IntegrationsAhead+8
Pricing and costBehind-6
ComparisonsBehind-13
Security and complianceBehind-19
MigrationBehind-27
BenchmarksBehind-34

Winning setup questions and losing pricing means you get cited to people who already chose you, and ignored by people still deciding.

What gets quoted

Pages that survive the filter share a shape. None of it is a trick. It is what makes a page useful to someone in a hurry.

Anatomy of a page an answer engine will quote A stylised article page with five annotated features: a definition in the first forty words, headings phrased as real questions, specifics held in a liftable table, a source attached to every claim, and a named author with credentials. Anatomy of a page an answer engine will quote A definition in the first 40 words, before any preamble Headings phrased as the questions buyers actually type Specifics in a table a model can lift without paraphrasing Every claim that needs evidence carries its source A named author whose credentials check out elsewhere
Five features that decide retrieval. Most pages that fail are missing the first one. Three hundred words of warm-up loses the passage that would have been quoted.

What we publish

Four formats, each earning citations in a different part of the question set.

01

Answer-shaped explainers

A question as the heading, a settled answer in the first paragraph, depth underneath. These get retrieved because the opening passage stands on its own when it is torn out of the page.

02

Original research

A survey, a benchmark, a dataset nobody else has published. A number with a method behind it has exactly one source, and naming it is the only safe way to use it.

03

Comparisons and alternatives

The pages buyers ask for by name and most companies refuse to write honestly. Handled fairly, they get cited because the fairness is legible.

04

Practitioner field guides

Long, structured, revised as the tooling moves. They carry the specifics that fill out an answer once the headline question is settled.

How we work

  1. Map the question set

    We build the prompts your buyers use from sales calls and support tickets, run them across six engines, and record who gets named.

  2. Report before outlines

    Two or three interviews with people who did the work. No outline gets written before those conversations happen.

  3. Write, then structure for retrieval

    Written for a person first. Then the definition up top, the schema, the table holding the specifics, the sources attached to claims.

  4. Re-run the prompts

    Six weeks after publishing we run the set again and report what moved, including the quarters where a piece did not land.

6 Answer engines tracked in every audit.
3 weeks Kickoff to coverage map, before any writing is commissioned.
2 rounds Revisions included on every piece.

We do not promise attributed revenue. Answer engines strip referrer data often enough that clean attribution is not available to anyone honest.

Find out who the machines name instead of you.

The audit runs three weeks, costs a fixed fee, and involves no writing. If the answer is that your existing pages need restructuring rather than replacing, that is cheaper and you should hear it first.

shaban@meliada.com