Who Handles AIO? Roles and Monthly Workload for Small Teams

AIO work (getting your company presented accurately in AI answers) divides into four roles: measurement, writing, fact-checking and site settings. One person can hold several, but we recommend that the person who writes not be the one who checks the facts.

This four-role split is our suggested approach, not an industry requirement.

“We have no specialist, but my manager asked me to start AIO.” For part-time owners in that position, this article covers how to divide roles, examples for teams of two or three, how to estimate monthly workload, what to outsource and what to decide in the first month. If you want the meaning of AIO first, read What Is AIO?

The four roles needed to run AIO in-house

We suggest dividing AIO work into four roles: measurement, writing, fact-checking and site settings. Each needs different access, so define roles before assigning people.

Role	Main tasks	Access needed	Can be combined with
Measurement	Record AI answers per question and review Google and Bing reports	View access to Search Console and Bing Webmaster Tools	Site settings, writing
Writing	Draft fixes for pages missing answers or conditions	CMS editing	Measurement
Fact-checking	Verify prices, conditions and scope	Ability to ask the department that sets terms and prices	Anything except writing
Site settings	Check that pages are usable in search and crawling is not blocked	Changes to robots.txt or server settings	Measurement

The site settings role exists because Google applies standard SEO basics to its AI features. Google lists checks such as allowing crawling in robots.txt and in CDNs (external services that deliver pages quickly), providing important content as text and matching structured data to visible text [1].

Separating the checker from the writer keeps “the text is finished” apart from “the conditions were verified.” Readable text still cannot be published if deadlines or prices are wrong.

Example role splits for teams of two or three

With few people, give writing and fact-checking to different people, and combine measurement with site settings. The team setups below are hypothetical.

Setup (hypothetical)	Person A	Person B	Person C
Two: marketing owner + web developer	Writing, requesting fact-checks	Measurement, site settings, fact-checking	—
Three: marketing owner + web developer + sales lead	Measurement, writing	Site settings	Fact-checking

With two people, A writes and B checks so no one reviews their own work. For prices or contract terms neither person can decide, ask the department that sets them.

With three people, the sales lead who explains real terms in meetings is well suited to fact-checking. They also know frequent customer questions and can help choose questions for measurement.

A weekly flow for two people from research to edits and approval is covered in Lessons from Plaid’s Small Team.

Monthly AIO workload: estimating measurement, fixes and checks

Monthly workload depends on the number of target pages and questions, so there is no fixed time. Break the work into small units, time one unit yourself and multiply.

Work unit	Counting unit	Your estimate	Example entry (hypothetical)
Test and record one question in one AI	Questions × AI services × tests	__ min	5 min
Copy report figures to the record	Once a month	__ min	20 min
Draft a fix for one page	Pages	__ min	60 min
Fact-check one page	Pages	__ min	30 min
Check settings	Once a month	__ min	30 min

Because AI answers change even for the same question, decide how many times to ask each question and multiply by that too. Using the example entries, 10 questions × 2 AI services × 3 tests is 300 minutes, and fixing and checking 2 pages is 180 minutes. This shows how to calculate; it is not an actual workload. Record the real time spent in the first month and use it for the next month’s estimate.

Monthly tasks for measurement: reviewing Search Console and Bing AI reports

The measurement owner records Google and Bing reports separately and does not mix them with clicks or visits.

Search Console’s Generative AI Performance Report is described as rolled out to all websites on August 31, 2026 [2]. It shows how often your URLs appeared (impressions) in features such as AI Overviews and AI Mode, in separate Search and Discover reports, by page, country and date, with a device breakdown for Search only [2]. Clicks are not among the announced items, so check visits separately in your analytics.

Bing Webmaster Tools’ Citation Share launched in preview on June 16, 2026 [3]. It is your site’s share of all citations across all sites for a specific grounding query, the search query the AI uses to find supporting sources [3]. It is separate from the answers the measurement owner records per question. A Compare feature for comparing with a previous period launched at the same time [3]. Microsoft describes Citation Share as an observational metric, not a ranking or traffic share [3].

Open the reports by selecting your site in Search Console and Bing Webmaster Tools. Both require a verified site and view access; if you lack access, ask your site administrator.

Set up these columns in your monthly record:

Month	Impressions in Google’s AI features	Pages with most impressions	Bing Citation Share	AI-referred visits (analytics)	Notes

Do not attribute a change straight to page edits. Keep the edit date alongside the record date, and note seasonality or changes in measurement.

Monthly tasks for writing and fact-checking: fixing one page at a time

The writer and checker pick one page and treat fixing and verifying its answer, conditions and evidence as one cycle.

1. From the measurement record, pick one question where your company did not appear.
2. Open the page that answers it. Check whether the answer is at the top and whether conditions are hidden far away.
3. The writer drafts the fix and attaches the terms or price list used as evidence.
4. The checker verifies deadlines, prices and scope against the evidence. Hold publication if anything is unverified.
5. After publishing, set a date for the measurement owner to recheck the same question.

Google says meeting its conditions does not guarantee inclusion in AI features [1]. So report “the page was fixed” separately from “the answer changed.”

Deciding what to keep in-house and what to outsource

Keep fact-checking of your own prices and terms in-house; recording results and checking site settings can be outsourced. The dividing line is whether your company holds final responsibility for accuracy.

Task	Outsourcing	Reason
Recording and totaling answers	Can outsource	Once questions and conditions are set, it follows a procedure
Checking site settings	Can outsource	Technical checks with few internal decisions
Drafting fixes	Partly	Drafts can be delegated, but your company supplies the evidence
Fact-checking and publishing decisions	Keep in-house	Terms can be shared with the vendor, but your company decides who gives final approval
Choosing questions to measure	Led in-house	Your company knows what customers actually ask

Even when outsourcing, name an internal owner for the question list and pre-publication checks. The proposal comparison in LLMO, GEO, AEO and AIO can also help you compare requests.

What to decide in the first month

You can start once you decide four things: owners’ names, three target pages, where the record lives and check dates.

• Owners’ names: write a real name for each of the four roles. Combining roles is fine.
• Three target pages: choose pages that answer customer questions, such as pricing or high-inquiry product pages.
• Where the record lives: somewhere everyone sees the same sheet, such as a shared spreadsheet.
• Check dates: put the monthly measurement date and post-fix recheck dates on the calendar.

In the first month, also record the time each task took. Next month, put those actuals into the estimate column and decide whether to add pages or questions. To keep recording answers per question, see How to Use AI Optimization for AgentSignal’s AI citation rate workflow.