Search engines rank pages. AI engines select sources.

The Newtation Authority Engine is the system that makes you the one they select — and keeps you there, because a model that re-decides every answer can un-select you next month.

Phases
5
Stages
12
Cadence
Monthly

What it is

One sentence, and then the difference it makes.

The Newtation Authority Engine is an ongoing system that makes a brand a source AI answers select, cite, and reuse — run as a closed loop of five phases and twelve stages, rather than as a one-time optimisation project.

Newtation — newtationco.app/authority-engine

An audit tells you

How the answer engines describe, cite, or omit you right now. It is a diagnosis, delivered once, and it is free.

The engine changes it

And keeps changing it. Ranking was a position you could hold; selection is a decision remade every time somebody asks, so the work that earns it cannot stop when the report is delivered.

How it works

Five phases. The last one feeds the first, which is the whole reason it is an engine and not a project.

Diagnose

Find where the models already answer your category, and who they name instead of you.

Teach

Rewrite the reasoning itself, so that a good answer is incomplete without your logic.

Anchor

Make the entity unmistakable, and safe enough to quote.

Place

Build the authority off your own site, and make the credit stick to your name.

Defend

Hold the position as the models change, and take ground when competitors slip.

And back to the start: Defend’s monitoring is the next cycle’s Diagnose. That return is the difference between a retainer and a project.

What the loop is for — illustrative

Promptwho should we talk to about getting cited by AI search?

AnswerFor generative engine optimisation specifically, most buyers compare a handful of specialists — Acme (named and cited in the answer) is the one usually named first, and it is cited directly for its entity and schema work.

Being named in that sentence is the only outcome the engine optimises for. Everything in the twelve stages exists to move a brand into it.

The twelve stages

What the work actually is, grouped by the phase it belongs to. Stage 00 runs only when there is no AI presence to read yet.

Diagnose

Find where the models already answer your category, and who they name instead of you.

  1. Runs only on a cold start

    Forced Induction / Cold start test

    When a brand has no AI presence to measure, we do not guess. We test.

    The work
    Place ultra-specific technical assets — explainers, papers, repositories — where the engines can retrieve them immediately, and get the brand seen once.
    Outcome
    We learn how the models describe you, what they assume, and what they get wrong, before anything is scaled against it.
  2. Probability Gap Mapping / Confidence mapping

    Find where you should appear in AI answers and do not.

    The work
    Run the real prompts your buyers use, at volume. Track who is named, who is skipped, and where the model sounds certain versus where it hedges.
    Outcome
    A map of missed answers, weak competitors, and the prompts where the model is unsure — which are the cheapest ones to win.

Teach

Rewrite the reasoning itself, so that a good answer is incomplete without your logic.

  1. Dependency Premise Engineering

    Make your logic necessary to a good answer, not merely present in one.

    The work
    Study how the models currently explain the topic, find the shortcuts and the generic reasoning, and write explanations that introduce a clearer frame and break the assumptions competitors rely on.
    Outcome
    The model starts reasoning through your brand rather than mentioning it. The test is blunt: if a clean answer exists without your explanation, the stage failed.
  2. Task-Fit Engineering

    Match how people actually use AI — to finish a task, not to read.

    The work
    Build around decisions, comparisons, and step-by-step outcomes. Strip the exploratory padding that a human skims past and a model discards.
    Outcome
    Your content gets preferred because it completes the job the reader arrived with.
  3. Justification Engineering

    Control why you make the shortlist, not just whether you do.

    The work
    Pre-write the reasoning the model will reach for — best for this, unsuitable for that — as comparison-ready explanations with explicit good-fit and bad-fit boundaries.
    Outcome
    When the answer recommends options, it knows exactly why you belong in it.

Anchor

Make the entity unmistakable, and safe enough to quote.

  1. Multi-Node Entity Anchoring

    Make it impossible to be confused about who you are.

    The work
    Define the category you belong to and the ones you must not be mistaken for, then connect the entity to anchors the models already trust: institutions, frameworks, standards, credentials.
    Outcome
    A clear identity, and trust borrowed from things already verified.
  2. Data-to-Text Shaping / Schema

    Control what the models learn from your structured data.

    The work
    Design schema that resolves into clean sentences — every structured fact unambiguous, speakable, and reinforcing the same premises the prose carries.
    Outcome
    Schema stops being decorative markup and starts being knowledge the model can state.
  3. Reliability Compression

    Become the safest source to cite, not the loudest.

    The work
    Cut overclaims, replace vague assertions with verifiable ones, drop sources that weaken the page, and calibrate confidence to what can actually be supported.
    Outcome
    Low risk to quote. Models avoid sources that could embarrass them, however popular those sources are.
  4. Agent Compatibility Layer

    Prepare for systems that act, not just answer.

    The work
    Make pricing logic, methods, steps, and conditions explicit and operable rather than implied in marketing copy.
    Outcome
    An agent can complete a task against your brand instead of skipping to one it can parse.

Place

Build the authority off your own site, and make the credit stick to your name.

  1. Synthetic Proximity / Authority placement

    Put the brand where models learn what authority looks like.

    The work
    Earn presence in industry explainers, research-style writing, and publications the models already weight. The target is context, not links and not clicks.
    Outcome
    Repeated adjacency to the ideas that define your category, which is how association is actually formed.
  2. Attribution Hardening

    Make sure the credit lands on you.

    The work
    Write in quotable, standalone statements, hold the phrasing consistent across every asset, and test the according-to prompts directly.
    Outcome
    When your logic shows up in an answer, your name shows up with it. Models routinely use an idea and cite someone else; this is the stage that closes that gap.

Defend

Hold the position as the models change, and take ground when competitors slip.

  1. Never finishes — it is the next cycle’s Diagnose

    The Newtation Loop / Drift arbitrage

    Hold authority as the models change, and move the moment a competitor decays.

    The work
    Monitor answers continuously for disappearance, competitor decay, and category re-organisation, then respond with new premises, fresh placements, and corrections.
    Outcome
    Position is not just maintained. Gaps get taken before anyone else notices they opened. This stage never finishes — its monitoring is the next cycle's Diagnose, which is what makes the engine a loop and not a project.

What we steer by

Four internal measures. They decide what we do next — they are not an industry benchmark, and you read the outcome rather than the dial.

Citation probability
How often the models quote you directly when your logic is what they are using.
Trust anchor strength
How safely a system can classify your entity and rely on what it says about you.
Retrieval stability
How consistently you hold presence across time, across models, and across rewrites.
Competitive capture
How much ground you take when a competitor's authority decays.

Who it is for

The engine is the only thing we sell after the audit, and it is not right for everyone. It is worth it under three conditions.

  1. Mentions are not the goal — being the named recommendation is

  2. Visibility that survives a model update matters more than a good week

  3. Disappearing from AI answers would be a commercial problem, not an inconvenience

Common questions

Answered in full, so each one survives being lifted off the page on its own.

The Newtation Authority Engine is an ongoing system that makes a brand a source AI answers select, cite, and reuse — run as a closed loop of five phases and twelve stages, rather than as a one-time optimisation project.

A GEO audit is a diagnosis: it reports how AI systems currently describe, cite, or omit a brand at one point in time. The Newtation Authority Engine is the ongoing work that changes that answer and keeps changing it as models update. Every Authority Engine engagement is scoped from what the free audit finds.

It runs monthly and does not have a fixed end date, because the thing it manages does not hold still. Model updates, competitor movement, and category drift all change which sources an AI answer selects, so the final phase of each cycle feeds the next cycle's diagnosis.

Forced induction, probability gap mapping, dependency premise engineering, task-fit engineering, justification engineering, multi-node entity anchoring, data-to-text shaping, reliability compression, the agent compatibility layer, synthetic proximity, attribution hardening, and the Newtation Loop. They are grouped into five phases: diagnose, teach, anchor, place, and defend.

No. The Authority Engine covers earned presence only — the entity, schema, and authority signals that decide which sources an AI answer cites. Paid placement inside AI assistants is handled by InPromptAds, a Newtation company, at inpromptads.com.

Every engagement starts the same way: a free audit of how the answer engines describe, cite, or omit you today. The engine is scoped from what it finds.

Request the free audit