White-label AEO report template: what to include

Section 1: How AI sees this website

One page, no jargon, before any scores. What can an engine currently establish about this business: what it does, where, for whom, on what evidence? Owners forward this page to their business partner. Write it so it survives that forwarding.

Section 2: Executive summary

Score, three biggest gaps, one-sentence plan. Written last, placed second. If the client reads only two pages, these are the two.

Section 3: Live tests, with the evidence attached

Each query gets its own table: the exact prompt, the engine, the date, what came back, whether the client was named, and which competitors were. Screenshot everything and date-stamp everything, one run per engine. The date-stamps matter because AI answers are measurably unstable: a study of 530,875 citations found most cited sources change day to day, with daily churn of 44 to 88% depending on the engine (GetMentions.ai, June 2026).

Put that volatility note in the client-facing report, with the citation. It reads as a confession and lands as a credential: you are the provider telling them what nobody honest can promise, which is exactly the contrast you want with whoever else is pitching them "guaranteed citations".

Section 4: Where the citations actually go

List the domains each engine cited on the client's queries. Semrush's analysis of 100 million+ citations shows AI answers lean on a recurring set of third-party platforms, weighted differently per engine (Semrush, 2025). Every cited domain where the client is absent is a concrete, explainable gap, and each one converts naturally into a line of follow-on work.

Section 5: The findings, scored in the open

Organise findings by audit dimension, each with its evidence and its severity. Then the score, with every weight printed beside its dimension. A score the client can recompute is a measurement. A score with hidden weights is marketing, and it cannot support a like-for-like re-test in 90 days. Resist the urge to pad the checklist: where the evidence says an item does not matter yet, say so. The llms.txt file is the canonical example, with 97% of published files never requested across a 137,210-domain study (Ahrefs, May 2026). Record it, cite that, and score nothing.

Sections 6 and 7: Fixes split by who does them

Two lists, never one. Copy-paste fixes the owner applies the same afternoon: rewritten H1, meta description, an FAQ block with the matching structured data. Developer tasks for everything needing access or expertise: schema installation, heading restructure, technical clean-up, each written so it can be handed to a developer verbatim. Mixing the two lists is how reports die in inboxes, because the owner cannot tell which lines are theirs.

Section 8: The 90-day plan, ending in a re-test

Prioritise by impact against effort into this week, this month, this quarter. Close with the re-test offer in writing: same queries, same engines, same weights, 90 days on. The evidenced first run makes the re-test provable, and the re-test is the recurring revenue.

How to write it so the client acts on it

The writing rules matter as much as the structure, because the reader is a business owner, not another consultant. Use the owner's nouns: the practice, the firm, the business, the crew. Put the plain-language sentence first and the technical name in brackets after it, "the block of code that tells AI this is a question and this is the answer (FAQ schema)", never the other way round. Keep sentences short. Quantify every finding you can, and date every number: an unsourced statistic in a client report is a liability with your logo on it. And write the fixes as instructions, not observations. "The homepage has no H1" is a note; "replace the homepage title with this line, provided below, ready to paste" is a deliverable.

Delivery: the PDF is half the product

Deliver the report as a PDF with your branding, then walk it on a 30-minute call. Walk the score, the live-test tables, and the this-week fixes, in that order, because owners act on fixes they can apply the same afternoon and momentum sells the rest of the plan. The call is also where the re-test gets booked: same queries, same weights, 90 days on, while the baseline is still on the screen. A report emailed without a call gets filed; a report walked through gets implemented, and implemented reports are the ones that generate referrals.

The white-label hygiene rules

A white-label template earns its keep on the discipline, not the branding slot:

• Practitioner notes live in marked blocks. Guidance for you ("what this finding means", "how to pitch the fix") sits in clearly delimited note blocks with a consistent marker, visually loud in the working copy.
• A print-safe system keeps notes out of the export. Styling that makes surviving notes impossible to miss beats memory. Never rely on remembering to delete.
• The final check is a search, not a read-through. Before exporting, search the whole document for the note marker. Zero hits or you are not done. A read-through misses one note per fifty pages; a search misses none.
• Cite your sources inside the client copy. Every market or research claim in the report should carry its source and verification date. If a claim cannot be traced to a primary source, leave it out, even when it is a number everyone repeats.
• Name engines factually. ChatGPT, Perplexity, Google AI Mode are the services you tested, nothing more. No implied affiliation, no logos.

What to leave out

No promised rankings. No "guaranteed citations". No screenshots of other clients' results presented as predictions. No unattributed statistics. The report that omits all four reads plainer than the competition's, and it is the one that still looks honest when the client re-reads it in six months, next to the re-test that proves what actually moved.