A defensible AI-visibility briefing format
A publication-ready format for summarising provider documentation, crawler and access changes, reproducible research, and measurement implications.
RankRift does not run an email list or send a briefing. This page is a format for writing your own AI-visibility change briefing, with the subscriber-data rules any list should follow.
In short
This briefing format is for people responsible for measuring how AI assistants describe a brand. An issue covers documented provider changes, shifts in crawler and access behaviour, reproducible research with its method attached, one measurement exercise, and a short source list.
Who this is for
- In-house SEO and growth leads who need to know when a model or crawler change invalidates their baseline
- Agency strategists who have to explain a sudden movement in a client's numbers
- Analytics and RevOps owners tracking how AI referral and agent traffic is identified
- Technical SEOs watching crawler user agents, access controls, and provider documentation
- Founders and marketing leads who want a periodic sanity check rather than a daily feed
When to use this
- You run a measurement programme and need a repeatable way to brief colleagues on provider changes
- You need a repeatable checklist for assessing a research run or dataset announcement
- You need a dated way to record the release of a template, checklist, or calculator
- You want a format whose default position is skepticism
What each issue contains
Use one compact issue on the chosen cadence. The organising question stays the same: did anything change that should alter how the reader measures or what the evidence supports?
When an item concerns model behaviour, link it to dated provider documentation or a reproducible run with its observation date, prompt set, model identifier, sample size, and variance. Label an unverified effect as unknown rather than filling the gap with speculation.
- Documented provider changes, linked and dated, with what they mean for measurement
- Crawler, access, and retrieval changes worth a look at your own configuration
- A research-evidence check covering prompt set, model versions, run dates, sample size, and variance
- One practical measurement idea you can apply in an hour
- A dated release note for any template, checklist, dataset, or calculator discussed
The recurring segments
Every issue uses the same structure, so you can skip the parts that are not your job. An issue with nothing to report in a segment omits it rather than padding it out.
- Changed and documented — provider updates with a link and a date, and a plain sentence on the measurement consequence
- Changed and unverified — things being widely repeated in the category that could not be confirmed, labelled as such
- From the runs — findings from your own measurement runs, always with the method attached
- One thing to try — a single, small, testable action, usually a change to how you sample or record
- Risk watch — platform policy, regulation, and manipulation tactics covered defensively
- Corrections — anything the previous issue got wrong, stated in the same place it appeared
Cadence, and what this is not
The template uses a fortnightly cadence because provider changes worth acting on do not usually justify a daily digest. A briefing using the template should state its actual cadence, identify any exception issue as such, and never use an alleged emergency as promotion.
- Not a daily digest of AI news, and not a feed of model release announcements
- Not a vendor press-release channel — it does not forward product launches, funding rounds, or feature announcements
- Not sponsored content dressed as editorial; any sponsor is labelled inline and has no say over selection, framing, or study results
- Not a drip sequence with a sales pitch at the end, and not a channel for guaranteed-outcome offers
- Not a place to publish a number without its sample size, models, dates, and variance
Subscriber-data requirements
An implementation should collect only an email address and, if justified, an optional first name. It should use double opt-in, send nothing before confirmation, and delete rather than retain an unconfirmed address after a stated interval.
A list operator should document any delivery or click data, its purpose, processors, retention period, deletion route, and whether tracking can be disabled. Subscriber data must not be sold, rented, or repurposed for public research. The list's own privacy notice, not this template, is the source of truth.
Unsubscribe requirements
Every issue should carry a working footer link that stops delivery without a retention call or confirmation gauntlet. Test the route when the list is set up and again after any change of email provider.
Deletion and suppression are different operations; the privacy notice should explain both and identify the request route. Editorial methodology and public reference material should remain available without subscription.
When this format is useful
Use this format for an audience that owns a measurement programme, presents numbers to skeptical decision-makers, or maintains the technical side of a site that assistants read.
Do not use it as a stream of product news, a sales sequence, or reassurance that a single lever changes how assistants describe a brand.
The research specification and measurement guide show the evidence standard in greater detail and are the right starting point before adapting this template.
Related pages
- Research study specification
- Methodology
- On-page measurement resources
- Privacy policy
- Editorial policy
Frequently asked questions
What cadence does the template use?
Fortnightly on a fixed day is the default. A briefing should state its actual cadence at signup and identify any exceptional issue. RankRift does not send a briefing of its own.
Should a briefing list run a sales sequence?
No. The format excludes sales sequences and requires separate consent for any different purpose. Public methodology and reference material should stay available without a subscription.
Should downloading a template add someone to the list?
No. Treat a download and a briefing as separate consents. A file request must not enroll the requester in a list; a separate, unticked choice and double opt-in are required for the briefing.
Should issues also be published on the web?
Yes. Keep the research and reference material cited by an issue on a durable public page, because a finding buried only in email is not citable. The email should filter and link rather than create an evidence paywall.
Can a briefing carry a sponsor?
RankRift takes no sponsorship. A publication using this format must label any sponsor inline, must not let a sponsor influence selection or framing, and must not let a sponsor be a subject of research it funds.
Why is there no subscriber count on this page?
Because RankRift runs no list. A subscriber count is social proof rather than evidence of usefulness, and a briefing should not imply a scale its list operator cannot document.
Apply the briefing format
Keep every item dated, sourced, and bounded to what the evidence supports.