Meta Ad Claims: How to Substantiate Every Number

In short: Every objective number in a Meta ad — a percentage, a comparison, a rating, an outcome timeline — needs a documented source you can produce before the ad runs, not after a complaint. A reliable operating principle is one file per claim: dated, with the exact wording, the method, and the conclusion it supports. What follows is Meta-platform-universal practice plus an internal evidence standard, not legal advice; consumer-advertising law varies by country, so confirm specifics with counsel or the regulator where the ad runs.

For the neighboring compliance checks, compare Children’s Product Meta Ads: Audience and Creative Compliance and use Condition-Related Food Meta Ads: Claims and Creative Compliance before approving the next campaign.

Scope and authority

Three different rulebooks govern a claim. Conflating them creates avoidable review and governance risk.

• Meta advertising policy is the platform authority. Meta’s published Advertising Standards (accessed 2026-08-27; paraphrased conservatively here) restrict deceptive claims, unrealistic outcomes, and certain personal-attribute assertions, and review can disapprove or restrict any ad. This surface is global and changes over time — re-check the live policy pages whenever you build creative.
• Your internal evidence standard is what you commit to yourself: no objective number ships without a named source and owner. Meta does not audit your files, but the register gives you a contemporaneous record to examine or produce when a claim is challenged — and it is the thing you fully control.
• Local consumer-protection law is the layer this guide cannot resolve. Whether a substantiation is sufficient, and what a regulator can do about a misleading one, depends on the advertiser’s jurisdiction. Treat cross-border claims as requiring local review.

When the three disagree, the strictest applicable rule governs.

Policy-risk map

Policy area	Risky execution	Review question to ask
Performance claims	A bare percentage or “3x” with no cited study	“Which document proves this exact number, and what did it measure?”
Comparative claims	“Better than the market leader,” method undisclosed	“Is the comparison tested, current, and stated so a reader knows what was compared?”
Superlatives	“#1” or “best-selling” with no basis	“Is there third-party ranking or a specific, verifiable scope?”
Ratings and review counts	A star rating and count frozen at launch	“Does the displayed rating match the live number today?”
Outcome and timeline claims	“Results in 7 days” as a universal promise	“Is there studied evidence for the timeline, and is variability disclosed?”

Claims evidence register

The register is the spine of the system: one row per active claim, so each claim has a documented route to its source. For each claim, record these fields.

1. Exact claim wording — verbatim from the ad, not a paraphrase.
2. Claim type — numeric, comparative, superlative, rating/review, or performance/outcome.
3. Source — the specific document: internal study, third-party lab report, certification, sales data, or public review platform.
4. Method and scope — how the number was produced: sample count (the denominator), conditions, comparison group, time window, statistical result. A claim is only as strong as its scope.
5. Conclusion the source supports — mapped to the exact claim, not an adjacent one. A study of “satisfaction” does not substantiate “effectiveness.”
6. Owner — the named person accountable for keeping the claim true.
7. Substantiated-on date and expiry — when it was verified and when to re-check. Ratings, review counts, and comparative claims can change over time; assign each its own re-check trigger.
8. Landing-page match — the destination page states the same claim with the same evidence, so ad and page do not diverge.

A useful internal rule is: if a row cannot be completed honestly, the claim does not run.

Scenario assumption vs fact

Numbers here are illustrative only. To show how a register entry reads, take a scenario assumption, not an industry benchmark or a Bach.ai result: suppose an internal test recorded that, of 200 surveyed users after 30 days, a share reported a visible improvement. The defensible ad claim mirrors that scope exactly — “based on a survey of 200 users after 30 days” — because the denominator, the window, and the method are all stated. A bare “visible results” is the same product but not traceable to a source. That distinction between a modelled assumption and a verified fact must survive into the copy unchanged.

Creative and copy pre-flight

Run this before any creative goes to review; it folds the register into a per-ad copy check.

• Personal attributes. Copy must not assert or imply knowledge of a viewer’s personal characteristics — health condition, financial situation, identity. Meta has published restrictions in this area, so check Meta’s current policy before relying on this; as an internal standard, write to a general audience.
• Prohibited implications. No implied cure, assured result, or “no side effects” absolute for anything applied or ingested — these are high-risk claims that require unusually strong evidence and may also conflict with applicable platform or local rules.
• Transformations and before/after. If you show change, disclose that conditions are identical and results are not universal; document material edits and reject any edit that changes the depicted product, condition, or result in a way the evidence does not support.
• Testimonials. Use only genuine, permissioned testimonials; document typicality and obtain local advice on required disclosures. A customer’s voice does not exempt an objective claim from needing its own evidence.
• Disclosures. As a conservative internal standard, place any qualifier a claim depends on (“based on 200 users,” “results vary”) prominently in the ad — not buried, and not only on the landing page; confirm any specific placement requirement under current platform policy and local law.
• Destination consistency. The claim, number, and qualifier on the ad must match the landing page; a mismatch is a policy risk.

Audience and data controls

Substantiation extends to targeting: if a claim implies you know something about the audience, the data behind it needs the same discipline as the claim.

• Permitted audience. Do not build or imply targeting on sensitive personal attributes; keep audience logic to behavior and interest signals — and check Meta’s current policy before relying on which signals are permitted.
• Source and collection assertion. For any custom audience built from your own data, record its origin, the asserted legal basis, and any notice or consent required in the applicable jurisdiction — that assertion is a register entry too.
• Suppression, retention, access. Keep a documented way to honor deletion and suppression requests, a retention limit, and a short list of who can access the source data.
• Jurisdictional review. Data-protection obligations differ sharply by country. Operating across borders, have local counsel confirm your audience-data practices — this guide cannot stand in for that review.

Approval workflow

A claim should pass through named hands, not a single rushed sign-off.

1. Drafter writes the copy and links each objective claim to its register row.
2. Evidence owner confirms the source supports the exact wording and that the substantiation is current, not expired.
3. Policy review checks the creative against the live Meta advertising policy and the pre-flight list above.
4. Local expert — counsel or a regulatory specialist — reviews anything category-sensitive or cross-border before launch. Build this in for regulated categories rather than bolting it on after a rejection.

If an ad is rejected or payment fails

Diagnose from what the account actually shows, and change one evidenced thing at a time.

• Read the stated reason first. Read whatever reason or policy area Meta surfaces; if it is generic, do not infer a more specific cause without evidence. Map it to a specific claim or creative element rather than guessing broadly.
• Fix one cause, resubmit, observe. If a claim was flagged, tighten that claim’s wording to match its evidence and resubmit only that change, so the change remains interpretable, while recognizing that a later approval does not prove which factor determined the review.
• Payment or billing stops. Check the account’s own status signals — payment method, spending limit, balance, verification prompt — before assuming a policy issue. Use the account’s surfaced status to distinguish possible billing and policy causes; do not assume the category from symptoms alone.
• No promises. Compliant-looking creative is not a assurance of approval or reinstatement; the review outcome is Meta’s to make. Document what you changed either way.

Can software help?

Bach.ai audits your connected Meta account against 100+ checks, ranks what it finds by estimated impact, and proposes specific fixes. It stays read-only until you approve a change, then executes the approved change on Meta; connected Google Ads data is used for intelligence only. Think of it as an automated audit layer that surfaces issues and proposed fixes for your review — not a replacement for your team’s judgment, and it does not generate your creative.

Common mistakes

• One study, many claims. A small trial about “satisfaction” gets stretched to cover “effectiveness,” “speed,” and “beats competitors.” Each claim needs its own specific evidence.
• Frozen ratings. A star rating and review count set once at launch and never refreshed, so the ad quietly overstates the live number.
• Method hidden. “Clinically tested” or “2x more effective” with no accessible record of what was tested, against what, and how large the sample was.
• Qualifier only on the landing page. The scope that makes a claim honest sits on the destination page while the ad shows the bare number.
• No owner, no expiry. Claims with no accountable person and no re-check date silently rot until a challenge exposes them.

FAQ

What counts as substantiation for a Meta ad claim?

A specific, dated source supporting the exact wording — an internal study with a stated sample and method, a third-party lab report, a recognized certification, sales data with defined scope, or an accessible public review source. It must match the claim’s exact scope, and you should be able to produce it before the ad runs.

How is a scenario assumption different from a verified fact in ad copy?

A verified fact is traceable to a source you can show, with a known method and denominator; a scenario assumption is an illustrative or modelled figure with no such backing. Only verified facts belong in objective claims — label illustrative numbers as assumptions and keep them out of copy a reader reads as measured performance.

Is compliant-looking copy enough to get an ad approved on Meta?

No. Meta review is the platform’s decision and can disapprove or restrict any ad, and policies change over time. An evidence register and the pre-flight checklist help teams identify unsupported claims before submission and preserve a record for later review, but no process can promise approval or reinstatement.

How frequently should I refresh a claim’s substantiation?

It depends on the claim, not a fixed calendar. Rating and review-count claims drift as reviews accumulate and should be checked before publication and after any material change in the underlying rating, count, product, or comparison, and comparative claims move as the category moves. Set a re-check date per register row, and refresh sooner if the product, formulation, or underlying data changes.

Can a customer testimonial substitute for evidence of a performance claim?

No. A testimonial is itself an advertising claim and does not prove an objective statement such as a percentage or comparison. Use only real, permissioned customers, document typicality and obtain qualified local advice on required disclosures, and keep any objective claim backed by its own source. Whether a testimonial is acceptable can also depend on local law, so seek qualified review for sensitive categories.