Most AEO monthly reports in 2026 are vanity dashboards, and a useful one contains exactly six things: prompt level citation share across ChatGPT, Perplexity, Google AI Mode and AI Overviews, Gemini, and Copilot; an inventory of which URLs got cited; referral sessions and leads from AI sources in GA4; competitor share of voice on the same prompt set; first party engine data from Search Console and Bing Webmaster Tools; and a work log of what shipped. Trackers like Otterly.ai, Profound, and the Semrush AI Toolkit run 15 to 500 prompts a month for $29 to $499, so the raw data is cheap. What is rare is a report that ties that data to a frozen prompt set, a baseline, and a lead count.
What sections does a useful AEO monthly report contain?
A useful report has six sections, in this order: citation share by prompt and engine, cited URL inventory, AI referral sessions and leads, competitor share of voice, first party engine data from Bing Webmaster Tools and Google Search Console, and the work log. Anything else is an appendix.
1. Prompt level citation share by engine
This is the core table. Take a fixed set of buyer prompts (“best car accident lawyer in Charlotte”, “best rhinoplasty surgeon in Scottsdale”), run them on a schedule across ChatGPT, Perplexity, Google AI Mode, Google AI Overviews, Gemini, and Copilot, and record whether the client was cited, named, or absent in each answer. Citation share is cited answers divided by total answers, per engine. The prompt set is frozen at kickoff; if prompts change month to month, the trend line is fiction.
2. Cited URL inventory
Which of your pages did the engines actually pull? Profound, Peec AI, and Scrunch AI all expose the URLs behind each answer. The report lists every client URL cited this month, by count and engine, plus the third party pages cited instead of the client: an Avvo profile, a Justia listing, a local news story, a RealSelf page. That second list is next month’s PR and content roadmap.
3. AI referral sessions and leads in GA4
GA4 segments AI referrals by source: chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai. Since May 13, 2026, GA4 also ships a native “AI Assistant” default channel group, though a custom channel built on a source regex still catches newer hosts. The report shows sessions, form fills, and calls from that channel, month over month. Statcounter’s April 2026 data puts ChatGPT at 76.85 percent of AI referral traffic, Gemini at 9.0, Perplexity at 7.73, and Copilot at 3.76. AI referrals are still around 1 percent of web visits by 2026 industry estimates, which is why leads, not sessions, are the number to watch.
4. Competitor share of voice on the same prompts
Share of voice is the percentage of answers in the prompt set that name each brand. Run it for the client and three to five competitors named at kickoff. Alone, a 12 percent citation share tells you nothing. Against competitors, you know whether 12 percent is the category lead or a distant fourth.
5. First party engine data
Bing Webmaster Tools launched an AI Performance report in public preview on February 9, 2026: total citations across Microsoft Copilot and Bing AI summaries, citation counts by URL, and sampled grounding queries (the reformulated queries Copilot ran before citing your page). Google Search Console announced a dedicated AI performance view on June 3, 2026, UK properties first; it reports impressions for AI Overviews and AI Mode but does not separate AI clicks from regular web clicks. Paste both screenshots. They are the only citation data that comes from the engines rather than a scraper.
6. The work log
Every page published, schema block deployed, press placement earned, review answered, and listing corrected, with dates and URLs. When sections 1 through 5 move, the work log explains why. When they do not, it shows whether the engines are lagging or nothing was done.
How many prompts should the tracker run, and what does it cost?
Fifty prompts per engine is the floor for a defensible share of voice number, a single location service business needs 50 to 150, and published tools cap plans between 15 and 500 prompts. Below 50, single prompt variance swamps the signal.
Published pricing as of September 2026, per vendor pricing pages:
- Otterly.ai: $29, $189, and $489 per month for 15, 100, and 400 prompts; $99 per extra 100.
- Profound: $99 per month for 50 prompts, $399 for 100, custom enterprise above that.
- Peec AI: from 89 euros per month.
- Semrush AI Toolkit: $99 per month for one domain and 25 prompts, about $60 per extra 50.
- Ahrefs Brand Radar: $199 per AI platform per month or $699 bundled, on top of a $129 base plan; built on a database of 458 million monthly AI prompts, so it shows demand rather than tracking your own list.
- Scrunch AI: Core at $250 per month for 125 prompts across ChatGPT, Perplexity, AI Overviews, and Copilot.
- Rankscale: tiers from $79 per month.
- AthenaHQ: free Essential tier; Starter at $295 per month across nine models.
- Goodie: entry estimates of $199 to $495 per month; enterprise is custom.
- SE Ranking: folds an AI visibility tracker into its rank tracking plans.
Otterly.ai and the Semrush toolkit center on prompt tracking and mentions. Profound, Peec AI, and Scrunch AI add cited URL analysis and competitor share of voice. Ahrefs Brand Radar leans on prompt demand.
The buyer’s math: raw tracking costs $100 to $500 a month, or 5 to 15 percent of a $2,000 to $6,000 retainer. The rest pays for the work log.
Before you sign anything, Get your free AI visibility audit and see which prompts your firm or practice is already cited for across ChatGPT, Perplexity, and Google AI Mode.
Which metrics should you ignore in an AEO report?
Ignore any number that has no prompt behind it. Four vanity metrics show up in most agency reports, and each can rise while your citation share falls.
Vanity 1. “AI mentions” with no prompt context
A count of 340 mentions means nothing without the prompts that produced them. Mentions on “who founded X” and “best DUI lawyer in Denver” are not the same asset. If the tool cannot show the prompt, the number is decoration.
Vanity 2. Sentiment scores with no baseline
Profound, AthenaHQ, and others score how positively an engine describes a brand. Useful only against a month zero baseline and competitors on the same prompts. A sentiment of 82 out of 100 with no comparison is a mood, not a metric.
Vanity 3. AI impressions
Search Console’s AI view reports impressions for AI Overviews and AI Mode. An impression as source nine of twelve with no click is not visibility a client would pay for. Impressions belong in an appendix.
Vanity 4. Blended traffic charts
“Organic traffic up 8 percent” folds AI referrals into a bucket dominated by classic search. GA4 can separate the AI channel; a report that does not is hiding whether the AI work produced anything. Discount “estimated AI reach” figures too: they are modeled from prompt volume guesses, not measured from answers.
How do you read month over month movement without fooling yourself?
A citation share move under about 10 percentage points on a 50 prompt set is noise until it holds for two consecutive months. A spring 2026 repeat answer study of 16,143 ChatGPT prompts and 15,805 Google AI Overviews prompts found two runs of the same ChatGPT prompt shared only 21.2 percent of cited domains; AI Overviews was steadier at 31.5 percent. A single run is a coin flip, not a measurement.
Five reading rules follow:
- Three runs per prompt minimum, averaged.
- Compare rolling three month windows, not this month against last.
- Require a second engine to confirm. A 15 point jump on Perplexity with ChatGPT flat is usually a model update, not your work.
- Weight by engine share. Five points on ChatGPT, at 76.85 percent of AI referrals, beats 15 points on Copilot at 3.76 percent.
- Match expectations to the timeline. Benchmarks published in May 2026 found half of new pages earn a first AI citation within 7 days and 90 percent within 37 days, but domain level share builds slowly: 5 to 15 percent on priority prompts after six months, above 30 percent for the strongest accounts at 12 to 18 months. Month one and two reports should show the work log and first URL citations.
The scoring framework behind these rules is in how to measure an AEO agency.
What does a sample AEO monthly report skeleton look like?
A good report fits on six pages, opens with the lead number, and ends with the work log. This is the skeleton Subscribe PR uses for law firm and cosmetic surgery clients; any agency can run it with a $99 to $250 tracker plus GA4.
Page 1. Summary. AI channel leads this month, last month, and month zero. Citation share by engine, three month rolling. One sentence on what moved and why.
Page 2. Prompt table. One row per frozen prompt. Columns: ChatGPT, Perplexity, AI Mode, AI Overviews, Gemini, Copilot. Cell value: cited, named, or absent, averaged across three runs.
Page 3. Competitor share of voice. Same prompts, client plus three to five competitors plus the top directory (Avvo, Justia, RealSelf). Three month trend per brand.
Page 4. Cited URL inventory. Client URLs cited, by count and engine. Third party URLs cited for the same prompts.
Page 5. First party and GA4. Bing Webmaster Tools AI Performance screenshot with top grounding queries. Search Console AI impressions, flagged as impressions only. GA4 AI channel: sessions, form fills, calls, by source.
Page 6. Work log and next month. Dated list of everything shipped with URLs, then three priorities tied to gaps on page 4.
This is what a buyer should expect from any retainer that includes AEO, including the tiers on our services page. If a proposal cannot commit to the format in writing, that is the answer.
What should you ask an agency whose report is missing these?
Ask six questions and expect a document in reply to each, not a call.
- What is the frozen prompt list, and can I have the file? No list means no baseline and no trend.
- How many runs per prompt per month, on which engines, with which tool? Otterly.ai, Profound, Semrush, or a spreadsheet: any answer works, no answer does not.
- Which of my URLs were cited this month, and which third party pages beat me?
- What is the AI channel in my GA4, and how many form fills or calls came from it? If they have not built the channel, they have never measured a lead from their own work.
- What shipped this month, with URLs and dates? Fewer than ten dated items on a multi thousand dollar retainer is a red flag.
- What is competitor share of voice on the same prompts, and who wins the ones I lose?
If the answers do not arrive, the fix belongs in the contract, not another meeting. The reporting clauses to write in are in AEO agency contract terms; the wider checklist is in how to vet an AEO agency.
Frequently asked questions
How often should an AEO agency report?
A written monthly report plus a live dashboard. Weekly written reports are noise: with ChatGPT sharing only 21.2 percent of cited domains between two runs of the same prompt, a weekly delta is mostly variance. Quarterly is too slow to catch a prompt set that stopped working. Monthly, with three month rolling trends inside it, matches how fast citation share moves on Profound or Otterly.ai.
Can Google Search Console show clicks from AI Overviews?
No. The AI performance view Google announced on June 3, 2026 reports impressions, pages, countries, and devices for AI Overviews and AI Mode, but clicks inside those surfaces stay blended into regular Web performance data, with no AI specific click, CTR, or query breakdown. Use GA4 referral sources for click and lead data and treat the Search Console AI view as a coverage signal.
Does Bing Webmaster Tools show Copilot citations?
Yes. The AI Performance report, in public preview since February 9, 2026, shows total citations across Microsoft Copilot and Bing AI summaries, daily average unique cited pages, citation counts by URL, and sampled grounding queries. Grounding queries are the reformulated searches Copilot ran before citing a page, the closest thing to first party prompt data any engine publishes.
What is a good AI citation share for a law firm or a cosmetic surgery practice?
Ask for competitor share on the same prompts before judging the number. Published 2026 benchmarks put a focused domain at 5 to 15 percent citation share on priority prompts after six months, with the strongest accounts above 30 percent at 12 to 18 months. A Scottsdale practice competing against RealSelf and three surgeons will sit lower than a solo firm in a small market.
Should I buy my own tracker or rely on the agency’s report?
Buy the cheapest plan that covers your prompt set and run it in parallel for the first quarter. Otterly.ai at $29 for 15 prompts or Profound at $99 for 50 is enough to verify the agency’s table. If the two diverge beyond the noise floor for two months, you have a conversation to have. If they match, drop the duplicate and keep GA4 as your independent check.
The takeaway
An AEO monthly report is a claim, and the six sections above are the evidence: a frozen prompt set scored across ChatGPT, Perplexity, Google AI Mode, AI Overviews, Gemini, and Copilot; the URLs that got cited and the pages that beat them; the GA4 AI channel with leads attached; competitor share of voice; the Bing Webmaster Tools and Search Console screenshots; and a dated work log. Tracking costs $29 to $499 a month, so the value of the report is never the data. It is the discipline of reading that data against a baseline. Start from a measured one instead of a dashboard: see exactly which prompts already cite your firm or practice, and hold the next report to that number.
Tagged