What a claims library is for
The approvals and compliance guide sets out the principle this tutorial builds on. Advertising rules in most markets share a core idea: marketing must not mislead, and claims must be capable of substantiationThe evidence a business holds, before a claim is published, that the claim is true; advertising rules in most markets expect claims to be capable of it. before they are made. In the UK the Advertising Standards Authority publishes the codes and rulings that apply; in the US the Federal Trade Commission's business guidance covers advertising, endorsements and related topics. That guide recommends three habits in a paragraph each: keep a substantiation file, maintain a claims library and flag claims at intake. This tutorial builds the first two and connects them to the third.
The library does three jobs. It makes reviewers faster, because they check a claim against an entry instead of rebuilding the evidence each time. It makes routine work safer, because builders take approved wording from one place rather than paraphrasing from memory. And it makes withdrawal possible, because a claim that stops holding can be found, marked and taken down everywhere it runs. The theory of constraints uses a pre-approved claims library as its example of getting more out of a constrained legal review: the reviewer spends time only on what is new. When finished work keeps coming back from that review, the diagnosis for legal rejecting marketing work shows whether a missing library, edited wording or another gap is behind the returns.
Several pages already put individual claims into the file. The guide to writing a brand tagline files a claim-based line before use, the guide to responding to competitors does the same for comparisons, and the prompt library tutorial keeps a claims block that should match the library. The guide to AI claims and consumer protection describes what one substantiation entry records. This tutorial builds the library they all point to.
Before you start
- Name the owner and the approvers. One person, usually in operations or marketing, owns the library's upkeep. Approval of each entry belongs to the reviewer the approval matrix names for claims, usually legal or compliance. Agree who does what with a RACI matrix: who finds evidence, who writes entries, who approves and who is told.
- Agree what the library is not. It records claims, evidence and approvals. It does not decide what evidence is adequate, interpret sector rules or replace legal advice. Health and treatment claims, for example, carry the extra care the healthcare marketing guide describes and always go to specialist review.
- Decide where it lives. A shared register with one row per claim, with links to the evidence, is enough to start. It should sit where the approval records and the asset library can link to it.
- Gather what exists. Any existing approved claims list, past legal review notes, test reports, certifications, customer research, product specifications and the brand guidelines' messaging section.
- Read the AI claims guide's entry description. The fields in this tutorial follow it, so entries made for AI-assisted copy and for everything else look the same.
The steps
Decide which claims belong in the library
Write down the scope. Most teams include every product or service claim, every comparison with a competitor or an earlier product, every statistic, every superlative, every claim about results customers can expect, every environmental or ethical claim, every testimonial or endorsement used in marketing and any claim-based tagline. Price and reference-price claims are usually handled through promotion review rather than the library; the sale promotion readiness checklist covers them and this tutorial does not interpret them. Write the scope down so builders know what to look for in their own copy.
Collect every claim already in live use
Start from what customers can see. Go through the website, product pages, packaging, ads, email templates and automated journeys, sales materials, social profiles and anything given to partners, and copy out every claim word for word with where it appears. Include the claims hiding in small places: review snippets used as Social ProofEvidence that other people have chosen, used or approved something, used to reduce a buyer's doubt., badges, award mentions, subject lines and creator captions. Collecting from live material, rather than from an old approved list, finds the claims that were paraphrased, strengthened or invented along the way, which are the ones most likely to lack evidence.
Separate objective claims from statements of opinion
Sort the collected claims. Objective claims say something that could be shown true or false: a comparison, a statistic, a test result, a feature, a "best" or "fastest". The approvals guide counts all of these as claims that need a substantiation entry. Other statements express the brand's view or style, such as a tone-of-voice phrase, and may not need evidence at all. The line between the two is not always obvious, and where a statement could be read as a measurable promise, treat it as objective or ask the reviewer. Testimonials and endorsements are a third group with their own requirements, covered in the next steps.
Find the evidence for each objective claim and date it
For each objective claim, find the evidence the business relies on: test reports, data extracts, certifications, customer research, the documented behaviour of a feature. Link it from the entry and record the date it was last checked. Where no evidence can be found, mark the claim as unsupported and stop using it in new work until the reviewer decides what to do; do not reword it into something vaguer and keep going. Whether the evidence found is enough for the claim as worded is for the reviewer, not the person compiling the library.
Write each entry with exact wording, conditions and markets
Each entry records the exact wording approved, what it means in plain terms, the evidence and its date, any conditions such as qualifying text that must appear nearby, the markets and channels where it is approved, who approved it and when, and where it is used. The table after these steps lists the fields. Exact wording matters: the AI claims guide notes how easily "helps reduce" becomes "eliminates" in a rewrite. For testimonials and endorsements, record the source and the relationship: the FTC's endorsement guidance, as the influencer contracts and disclosure guide explains it, expects a material connection between an endorser and the brand to be disclosed clearly and conspicuously, and the AI claims guide adds that endorsements must be genuine. Customer content reused as UGCContent about a brand made by customers or creators rather than by the brand itself. needs the same record.
Get each entry approved by the named reviewer
Send entries to the reviewer the approval matrix names for claims, in batches grouped by product or theme so the reviewer sees related claims together. The reviewer approves the wording as written, approves it with conditions, asks for changes or rejects it. Record the decision, the reviewer's name and the date on the entry. Only approved entries can be reused without fresh review. Claims that are already live and are rejected or sent back go straight to the withdrawal process in the final step, rather than waiting for the rest of the library to be finished.
Connect the library to intake and review
Add a claims question to the intake form, as the approvals guide recommends: does this work contain a claim, and is it from the library? Work that uses only approved wording, within its conditions and markets, can skip legal review and go through brand review and the launch check. Work with a new, changed, stronger or broader claim goes to the reviewer before build, as the campaign workflow guide places the gate. Give builders read access to the library and the creative and copy templates a link to it. For AI-assisted copy, the accuracy and review workflows guide checks generated claims against the file, and the prompt library's claims block should be updated whenever the library changes.
Set a recheck date for the evidence behind each claim
Evidence ages. A comparison becomes untrue when a competitor changes its product; a statistic dates; a feature is redesigned; a certification lapses. Give each entry a recheck date chosen by its owner and the reviewer for that kind of claim, and run a routine that lists entries approaching their date. The approvals guide names a claims audit among its periodic audits, reviewing the file for evidence that has aged and claims that are no longer true; the recheck dates are how that audit is run in practice, while the tutorial on auditing marketing approvals checks that live work actually used library wording or went to review. Product changes, new research and complaints are triggers to recheck sooner. Evidence on file is not the same as an experience that matches the claim, and the tutorial on checking marketing promises against the experience tests each claim, and the promises customers infer, at the step where customers find out whether it holds.
Retire claims that no longer hold and find where they run
When evidence no longer supports a claim, or the product, market or approval changes, mark the entry as retired with the date and the reason, and tell everyone who uses the library. Then find every place the claim runs and take it down, using the tutorial on withdrawing an asset from every channel, which covers ads, the site, email templates, automated journeys, partners and print. Affiliates keep quoting old claims long after they are retired, and the tutorial on checking affiliate content compares what partners publish with the library. Keep retired entries in the library, marked as retired, so nobody re-approves the same wording without knowing its history. If a complaint or challenge triggered the retirement, the decision about anything more belongs to legal advisers.
What each entry records
The fields follow the substantiation entry the AI claims guide describes, with the operational fields a library needs to be searched, reused and retired.
| Field | What it holds |
|---|---|
| Claim ID | A short reference used in briefs, approval records and the asset library |
| Exact wording | The approved text, word for word, including any variants approved separately |
| Plain meaning | What the claim says in ordinary language, so reviewers can spot stronger rewrites |
| Type | Objective claim, comparison, statistic, testimonial or endorsement, tagline, environmental claim |
| Evidence | Links to the reports, data, certifications or research relied on |
| Evidence last checked | The date someone last confirmed the evidence still supports the claim |
| Recheck date | When the evidence must next be checked |
| Conditions | Qualifying text that must appear, formats or contexts it may not be used in |
| Markets and channels | Where the claim is approved, and where it is not |
| Endorsement record | For testimonials and endorsements: the source, the relationship and the disclosure required |
| Approved by and date | The named reviewer and the date of approval |
| Status | Approved, approved with conditions, under review, unsupported, retired |
| Used in | The assets and work items that carry the claim |
If the team wants a KPIA metric tied directly to an objective. for the library, the share of claims in live work that match an approved entry, and the share of entries within their recheck date, say whether it is being used and kept current. The "used in" field is also where to start when a changed claim was updated on some pages and not others, which the diagnosis of conflicting information across your pages traces alongside the other routes by which pages come to disagree.
A pet food brand builds its library (illustrative)
Gonvexa Pet Foods builds its claims library
Gonvexa Pet Foods, an invented pet food brand selling dry and wet food for dogs and cats online and through pet shops, sent every pack, ad and email to an external legal adviser. Reviews took about 8 working days, and the marketing team could not say which claims had been approved before.
The marketing operations lead collected every claim from 46 pack designs, the website, 3 months of ads and the email templates, and found 112 distinct claims. After merging near-duplicates and sorting, 61 were objective claims, 9 were testimonials and the rest were tone-of-voice phrases the adviser agreed needed no evidence.
The product team found dated evidence for 54 of the objective claims: nutritional analyses, feeding trial reports and ingredient specifications. 7 had none. The most visible was "recommended by vets", which had been on the website for 2 years and traced back to a single conversation nobody had recorded. It was marked unsupported, removed from new work and sent to the adviser with the other 6. Each of the 54 supported claims got an entry with exact wording, conditions and the markets where it was approved; 3 were approved only for the UK because the evidence came from UK trials.
The adviser approved the entries in 4 batches. The intake form gained a claims question, and over the next quarter about two-thirds of work used library wording only and skipped legal review, while new and stronger claims still went to the adviser. Review time for those fell to 4 days, because the adviser was no longer reading routine copy. The "recommended by vets" line was retired and taken down from the site, 2 packs at their next print run and an affiliate's product listing.
Verify it worked
- Pick one live page or pack. Every claim on it is in the register in exact wording, with where it is used.
- Open a few objective-claim entries. Each links its evidence, shows the date last checked and names the approver and date.
- Find the unsupported claims. Each is marked as such and left out of new work, not reworded into something vaguer.
- Trace recent work through intake. Work using approved wording only skips legal review; a stronger claim goes to the reviewer before build.
- Scan the entries for recheck dates. Retired entries stay listed with a date and a reason.
If it goes wrong
- Reviewers still receive routine copy. Intake probably does not check against the library. Add the claims question to the intake form and give builders read access.
- A live claim has no entry. The library was probably built from an old approved list. Collect again from live material, then send the claim to the reviewer.
- No evidence can be found for a claim. It was probably paraphrased or strengthened along the way. Mark it unsupported and send it to the reviewer.
- An entry has no evidence date. Nobody can tell whether it still holds. Add the last-checked date and agree a recheck date with the reviewer.
- A retired claim still appears somewhere. The search for where it runs was probably incomplete. Use the "used in" field, then check partners and affiliates.
Next steps
- Read the guide to auditing marketing approvals to test the library against finished work.
- Read the guide to AI claims and consumer protection to compare your fields with the substantiation entry it describes.
- Read the guide to responding to competitors before adding comparison claims.
Common mistakes
- Starting from an old approved list instead of live material. The claims customers actually see have often drifted from what was approved.
- Paraphrasing approved claims. A changed word is a new claim. Builders take wording from the library, not from memory.
- Rewording an unsupported claim into something vaguer. The vaguer claim still needs evidence; send it to the reviewer instead.
- Deciding evidence is adequate without the reviewer. The library records evidence; whether it is enough is a judgement for legal or compliance.
- Entries without dates. Evidence with no check date cannot be rechecked, and nobody knows whether it still holds.
- Treating testimonials as copy. Endorsements need the source, the relationship and the disclosure recorded, as the influencer contracts guide explains.
- Deleting retired claims. Keep them marked as retired so the same wording is not approved again without its history.
- Building the library and not connecting it to intake. If work does not check against it at the start, it saves no review time. The governance hub shows where the library sits beside the approval matrix, and the approvals guide's claims audit keeps it honest.
