In the 30 days to 23 October 2026, PaidSync customers handed their AI assistants about 34,000 multi-step ad-account tasks, roughly 8,000 a week, and most of that work was reading and checking accounts, not changing them. This is the first monthly report built from PaidSync's own anonymised usage. Claude is the assistant three in four advertisers use, reporting is the main job, and only 29% of active accounts made a change at all.
Each finding gives the count, then what it means for anyone building with AI on ad accounts.
149 of 198 accounts (75%) that connected an outside AI assistant used Claude or Claude Code. 45 (23%) used ChatGPT, Codex or another OpenAI tool. Accounts can use more than one, and 77 did.
What it means. If you build workflows for one assistant, build them for Claude first, and test them in the others.
117 accounts (59% of those using an outside assistant) used Claude Code, against 86 (43%) using Claude chat and 32 (16%) using ChatGPT.
What it means. The person asking the AI to work on ad accounts is often technical or on a team that works in code, so scripted, repeatable checks suit them better than one-off chat questions.
157 of 203 accounts (77%) had a performance, query or reporting tool among their five most-used tools on at least one day.
What it means. The first thing advertisers hand to AI is getting answers out of their data. Writing the reports and checks is where the time saving is.
87 of 203 accounts (43%) had a tracking, GA4, Tag Manager or conversion-check tool among their daily top five, against 63 (31%) for wasted spend and search terms.
What it means. Advertisers verify that measurement works before they trust the spend numbers, so a tracking check is a first step, not an extra.
Of 74 accounts that made any change through PaidSync, 42 (57%) added negative keywords and 21 (28%) changed a campaign budget. Adding or creating things (keywords, ads, ad groups) was the most common change, in 58 accounts (78%).
What it means. Advertisers delegate the careful, reversible work first. Budget moves stay mostly with people.
74 of 256 active accounts made at least one change in the 30 days. Together they made 11,280 changes, 59% of them in Google Ads and 15% in Meta Ads.
What it means. Read access is the entry point. Trust is built by showing accurate numbers before anyone lets AI touch an account.
Of 207 accounts that used a connected platform, 117 (57%) used two or more and 83 (40%) used three or more. Google Ads was used by 152 (73%), Meta by 111 (54%), LinkedIn by 40 (19%) and TikTok by 31 (15%). Platforms with fewer than 10 accounts are not shown.
What it means. An AI assistant that only sees one ad platform answers half the question for most advertisers.
Across 176 accounts, the top 10% (18 accounts) made 78% of the 69,556 assistant tool calls. The typical account made 16 of them in 30 days.
What it means. AI use in paid media is still early for most teams and very heavy for a few. Expect the median to move as the heavy users' habits spread.
5.0% of assistant tool calls failed (3,467 of 69,556) and 1.4% of multi-step runs (493 of 34,222). We report this as context, not as a headline.
It does not report how often a person approves an AI suggestion, because the warehouse does not record it. It does not report which tasks run by name beyond the categories above, or results such as savings or ROAS.
149 of 198 accounts (75%) that connected an outside AI assistant used Claude or Claude Code. 45 (23%) used ChatGPT, Codex or another OpenAI tool. Accounts can use more than one, and 77 did. If you build workflows for one assistant, build them for Claude first, and test them in the others.
117 accounts (59% of those using an outside assistant) used Claude Code, against 86 (43%) using Claude chat and 32 (16%) using ChatGPT. The person asking the AI to work on ad accounts is often technical or on a team that works in code, so scripted, repeatable checks suit them better than one-off chat questions.
157 of 203 accounts (77%) had a performance, query or reporting tool among their five most-used tools on at least one day. The first thing advertisers hand to AI is getting answers out of their data. Writing the reports and checks is where the time saving is.
87 of 203 accounts (43%) had a tracking, GA4, Tag Manager or conversion-check tool among their daily top five, against 63 (31%) for wasted spend and search terms. Advertisers verify that measurement works before they trust the spend numbers, so a tracking check is a first step, not an extra.
Of 74 accounts that made any change through PaidSync, 42 (57%) added negative keywords and 21 (28%) changed a campaign budget. Adding or creating things (keywords, ads, ad groups) was the most common change, in 58 accounts (78%). Advertisers delegate the careful, reversible work first. Budget moves stay mostly with people.
74 of 256 active accounts made at least one change in the 30 days. Together they made 11,280 changes, 59% of them in Google Ads and 15% in Meta Ads. Read access is the entry point. Trust is built by showing accurate numbers before anyone lets AI touch an account.
Of 207 accounts that used a connected platform, 117 (57%) used two or more and 83 (40%) used three or more. Google Ads was used by 152 (73%), Meta by 111 (54%), LinkedIn by 40 (19%) and TikTok by 31 (15%). Platforms with fewer than 10 accounts are not shown. An AI assistant that only sees one ad platform answers half the question for most advertisers.
Across 176 accounts, the top 10% (18 accounts) made 78% of the 69,556 assistant tool calls. The typical account made 16 of them in 30 days. AI use in paid media is still early for most teams and very heavy for a few. Expect the median to move as the heavy users' habits spread.
PaidSync.ai runs 630+ tools across 14 ad and analytics platforms. Connect your accounts once and use them in the AI assistant you already work in.
Get Started Free Book a Demo