About RankRift
RankRift is a new, independent publication about measuring whether AI assistants recommend a brand. This page states what we are, what the name means, what we publish, how we work, and the things we will never publish.
Read the editorial policy · Read the disclosure
RankRift publishes vendor-neutral evaluation criteria for AI-visibility trackers. It does not operate or sell one, and the disclosure page sets out the rules that keep tool coverage independent.
In short
RankRift publishes methodology and education about AI visibility: whether assistants name, cite, and recommend a brand when a buyer asks for the best option. The site publishes the ruler rather than inventing readings: metric definitions, prompt-set guidance, model-version fields, sample-size planning, run-date fields, and variance rules accompany the manual procedure. RankRift does not operate or sell a tracker, and it does not claim a measured track record.
Who this is for
- Readers deciding whether a measurement source is worth citing in an internal deck
- Practitioners checking whether our incentives compromise the research before they rely on it
- Journalists and analysts looking for who stands behind a published figure
- Readers who want to know what RankRift publishes and what it refuses to claim
- Anyone who arrived via the brand name and wants to know what this site actually is
What RankRift is
RankRift is an independent publication focused on one question: when someone asks an AI assistant for the best option in a category, which brands get named, cited, and recommended?
That question has no trustworthy ruler today. Assistant outputs vary by model, phrasing, region, personalisation, and date, which means a single answer is an anecdote rather than a measurement. Most of the confident numbers circulating in this category come from companies that will not publish how they produced them.
So RankRift publishes the method rather than a measurement claim. A conforming result carries its prompt set, model identifiers, run dates, sample size, variance, method version, and changelog, giving a reader what is needed to re-run and challenge it.
What the name means
The rift is the distance between where a brand ranks in classic search and where it gets recommended by an assistant. Those two systems do not agree, and the gap between them is measurable.
A brand can sit high on a search results page and be entirely absent from the short answer a buyer actually reads. The reverse happens too. Naming the gap is the first step to measuring it, and measuring it is the first step to deciding whether it deserves any of your budget.
What we publish, and what we do not claim
RankRift publishes a vendor-neutral tracker specification: requirements for versioned prompt sets, repeated multi-model runs, share-of-voice and citation-source reporting, drift alerts, and open exports. It is an evaluation brief, not a product offer; RankRift does not sell a subscription tracker.
RankRift does not sell placement, recommendation, or any outcome inside an AI answer. It does not accept payment to change an evaluation, include a brand in an index, or alter ordering. RankRift has no advertising, sponsorship, affiliate, product-sale, or commissioned-research relationship.
The methodology, glossary, study specification, and manual check procedure are complete on-page references without an email gate.
Independence and honesty commitments
RankRift writes about trackers but does not operate or sell one. The disclosure policy excludes undisclosed commercial influence from coverage and requires affected material to be removed when a conflict cannot be managed.
- We never place ourselves first in our own comparisons, and we always name who a different tool or approach serves better.
- We never affiliate-link a competing product, because that conflict cannot be resolved by disclosure alone.
- We never describe a competitor's features, pricing, or accuracy from marketing material; anything we state is verified hands-on or left unstated.
- A measured visibility figure is a factual statement about model output, never a judgement about a company's quality.
- Disputes are re-run, not argued, and material corrections are logged with a date.
- Sponsorship is not part of the current publication model; paid influence over prompt sets, categories, results, or ordering is prohibited.
How we work
Measurement uses official model APIs only. We do not scrape assistant web interfaces, evade rate limits, or automate consumer accounts, and we do not describe how to do any of those things. That is partly an ethics position and partly a defensibility one: a number produced by a method that violates a provider's terms is a number you cannot cite in a board meeting.
Every claim about how a model behaves must link to dated provider documentation or a reproducible published run with its observation date, prompt set, model identifier, sample size, and variance. Unsupported behavior claims are omitted.
AI assistance may be used in drafting. The public admission rules are the enforceable standard: sources must resolve, numerical claims must carry their method, and legal or regulatory material remains general information tied to a current official source rather than a case-specific conclusion.
A new brand, stated plainly
RankRift is new. We have no client roster to show you, no awards to list, no case studies, and no user counts, so we are not going to invent any. Where a page could carry a number we have not measured, we describe the thing qualitatively and say why the number is missing.
Confidence in this category should be earned by showing the method, not by asserting expertise. Judge this publication on whether the methodology is checkable and whether any cited external research is genuinely reproducible.
About the name
RankRift is a marketing-analytics brand covering AI visibility measurement. Names get reused across the internet, and we are unaffiliated with any similarly named product, service, company, or account anywhere else. If you arrived here expecting something other than independent measurement guidance for marketers, this is not that.
Anything published under this name outside this domain is outside the scope of this site. RankRift does not run a contact inbox; the contact page explains where each kind of question is answered.
What we will never do
This list is short and non-negotiable.
- Guarantee a placement, recommendation, ranking, traffic, or revenue outcome, or imply we can control what a model says.
- Publish manipulation instructions. Astroturfing, review fraud, parasite pages, and prompt injection are covered defensively — what they are, how they are detected, what they cost.
- Invent a statistic, a sample size, a study, a testimonial, or a track record.
- Publish a measurement without its prompt set, model versions, run dates, sample size, and variance.
- Round an under-powered result into a headline, or quietly leave a stale page standing after it fails re-verification.
Related pages
- Editorial policy and independence
- Disclosure and conflicts
- Corrections and disputes
- Methodology
- Contact
Frequently asked questions
Who writes RankRift?
Pieces use a house byline unless a named author has relevant, verifiable expertise. RankRift does not attach invented credentials to a byline. Drafting may be AI-assisted; every admitted factual or numerical claim still has to meet the source and reproducibility rules above. Sensitive topics remain bounded to general information and current primary sources.
How does RankRift make money?
RankRift operates as an independent publication and does not sell a tracker, commissioned research run, placement, or measured result. The methodology, glossary, manual snapshot, and evaluation guidance are ungated. The current publication model excludes sponsorship and affiliate relationships from measurement coverage.
Why publish the methodology instead of keeping it proprietary?
Because a number nobody can reproduce is not evidence, it is marketing. In a category built on non-deterministic systems, publishing the method lets skeptics re-run the work and disagree in public. It also gives readers a standard they can apply to every vendor.
Can a study using this specification name brands?
Yes, when the study records which brands appeared in sampled answers. That is a statement about model output under stated conditions, not product quality. Publish the full method and provide a correction and re-run record for a disputed result.
Can I cite RankRift research?
RankRift currently publishes a study specification, not a results dataset. Do not cite RankRift for a measured result that is not present. When evaluating an external run, cite that run's dated data, prompt set, model identifiers, method version, and licence rather than citing this site generally.
What should happen when a page is wrong?
Append a dated correction in place with a short description of what changed and add a material correction to the correction record. Resolve a disputed measurement by re-running its recorded method rather than negotiating the interpretation.
Check the evidence before relying on it
The methodology and editorial policy are the two pages that should decide whether this site is worth your trust. Both are ungated, and both state the rules any number on this site has to meet.