What Does an AI Search Optimization Agency Actually Do Each Month?
An AI search optimization agency's monthly work is five jobs run together: tracking a fixed list of buyer questions across AI engines, fixing the site's…
TL;DR: An AI search optimization agency's monthly work is five jobs run together: tracking a fixed list of buyer questions across AI engines, fixing the site's technical foundations, writing new pages, improving existing ones and earning mentions on the sites those engines read. In an AI search retainer, month one is mostly foundation work and a baseline, while months two and three move the hours to new pages, page improvements and off-site mentions, with the same buyer questions retested. A retainer that only tracks prompts and sends a report is monitoring, and eSEOspace, an agency that sells retainers, concedes that retainers can "drift into reports without work". Strategi's Growth plan checks 100 buyer questions on 6 engines every day and each month ships 15 blog posts, 5 money pages (service, product, comparison or location pages), 6 improved pages and every technical fix its monthly audit finds. The practical test for any agency is to ask for its work log, not its report.
An AI search optimization retainer pays for five jobs, not a report
Generative engine optimization (GEO) is the practice of getting a business found, quoted and recommended by AI engines such as ChatGPT, Gemini, Perplexity and Google AI Overviews. Answer engine optimization (AEO) is the same work with the stress on direct answers, and vendors draw the line between the two names in different places. Strategi treats SEO, AEO and GEO as one job, because selling them as three services is a pricing decision rather than three kinds of work.
The job breaks into five recurring tasks. Measurement tracks a fixed list of buyer questions across the engines in scope, while technical work keeps the site reachable and readable. New pages answer questions the site does not answer yet, improvements fix pages that already exist, and off-site work earns mentions on the reviews, directories, forums and publications that engines read. Only the first task watches; the other four change something.
Google says its AI Overviews and AI Mode are rooted in its core Search ranking and quality systems, so standard SEO foundations still apply there. Ranking on Google is not the same as being cited, though. In our model a page passes two gates: retrieval (can the engine reach, fetch and render it) and selection (of the pages retrieved, which ones answer the background searches the engine runs). Query fan-out is the set of related searches an AI engine runs behind one buyer question, so a content gap is often a missing answer to a question the buyer never typed.
The months below follow that work on one anonymised account. Strategi does not publish client names, categories or results, so the log records what was done, not what it produced.
- Work log
- the dated record of what an agency changed in a month, on the site and off it, with a link to each change.
Month one checks whether AI engines and Google can see the site at all
Month one is foundation work: crawler access, technical fixes, consistent business facts and a baseline. At Strategi, work starts in week one, and the first pages and fixes go live before the month ends.
Can AI crawlers reach the pages?
AI companies split crawling across several bots, and OpenAI, Anthropic and Perplexity each run more than one. OpenAI says sites that block OAI-SearchBot, its search crawler, won't be shown in ChatGPT search answers, even though GPTBot, its training crawler, is a separate bot. A site that blocked every AI crawler to opt out of training has shut out the search crawlers too, and the fix is a robots.txt edit that belongs in week one.
The same week covers the rest of retrieval: pages that are indexed, render cleanly and carry their key facts in text rather than images or PDFs. Business facts get unified, so the name, description, category and pricing match across the site, Google Business Profile, LinkedIn and the directories that list the business. Schema is kept accurate for rich results, and we add an llms.txt file because it is harmless, though we never sell it as a tactic.
The baseline fixes the questions before any page changes
A buyer question is one distinct need a buyer puts to an AI assistant, tracked on a fixed list so each month compares with the last. The Pedowitz Group's AEO guide suggests a manual baseline: put 20 to 30 category questions to ChatGPT, Claude and Perplexity and score how often the brand appears. Strategi's plans track 50, 100 or 250 buyer questions on 3, 6 or 9 engines and check every question once a day, because one reading of an AI answer is weather, not climate.
Brand mention rate is the share of tracked AI answers that name the brand, and the baseline splits it into branded and non-branded questions. Being named when a buyer asks about the problem, without knowing the brand, is what reaches new buyers. Search Console, GA4, Bing Webmaster Tools and Microsoft Clarity get connected in the same month, and the baseline is saved before any page changes, so later wins can be checked against it.
Month one's log entries
GEO Catalyst's three-month retainer model closes month one with an approved operating plan: the prompt panel validated, the baseline run, the work queue finalised and owners confirmed. Strategi's month-one log adds the fixes shipped and the first pages sent for review. Each entry carries a date and a link, so the client can open the change instead of taking a report's word for it.
After month one, the hours move to pages and mentions
From month two, setup work shrinks and most of the hours go to new pages, improved pages and, on plans that include it, off-site work. Each month runs as a loop: read the question data, pick the gaps, ship the work and log each change so month three can retest against it.
New pages answer questions the site does not answer yet
New pages include money pages, not only blog posts. Strategi's Growth plan writes 15 blog posts and 5 money pages a month, and the Scale plan writes 20 and 10. Page count is not the goal: Google states that producing separate content for every fan-out variation primarily to manipulate generative AI responses violates its scaled content abuse spam policy. Strategi has seen a large D2C brand's roughly 400 programmatic pages lose about 60% of their AI citations month on month.
What an improvement pass changes on an existing page
Improving what exists usually ships faster than writing from scratch, because the page is already indexed; Growth improves 6 existing pages a month and Scale improves 20. Each pass states the answer near the top, swaps vague claims for numbers, ranges, conditions and stated limits, updates stale facts, and merges or prunes pages that compete for the same question. Google says there is no requirement to chunk content or rewrite it specifically for its AI features, so these changes are made for the reader first. Vague content becomes background; specific content gets quoted.
Off-site work earns mentions on the sites engines read
Strategi's view is that third-party mentions in reviews, listings, forums and the press track whether a brand gets named far more closely than domain authority does. Only the Scale plan works off the website: listings, Google Business Profile, directories, review collection, Reddit and Quora, 50 publisher pitches and an original data study each month. Google warns that chasing inauthentic mentions across the web is less effective than it appears, which is why the data study matters: it gives publishers something real to cite.
Month three retests the same questions
Month three keeps shipping pages and adds the comparison: the same buyer questions, on the same engines, read against the baseline. A regression sweep runs alongside it, because a redesign or plugin update can quietly undo a month-one fix.
Is a generative engine optimization retainer just prompt tracking and a report?
A generative engine optimization retainer can be exactly that: some are monitoring-only by design. GEO Catalyst, which sells software to agencies, sketches a monitoring package of 20 to 30 frozen prompts and a monthly report, with implementation scoped separately, beside an "Improve" package that adds 4 to 8 capped implementation hours and a monthly retest. GEO Catalyst calls these illustrative planning examples, not verified market averages. A monitoring retainer is honest when it is labelled as one; the trouble starts when a monitoring fee is sold as if it bought the work.
Strategi's pricing survey of 55 providers, checked 24 Oct 2026, covered 20 AI visibility tools, and they track: none of them writes a page, fixes a site or earns a mention. A dashboard can show where a brand is missing; it cannot write the missing page. For retainers that promise the work, NUOPTIMA recommends monthly acceptance criteria in the statement of work (pages shipped, prompts checked, citations reviewed, fixes submitted) plus a blocker log recording who held each item up.
| What to compare | Report-only retainer | Retainer that does the work |
|---|---|---|
| Monthly output | A visibility report, often with screenshots of AI answers | Pages written and improved, fixes shipped and mentions earned |
| Who makes the changes | The client's team, from a recommendations list | The agency, with the client's approver signing off before anything goes live |
| When the data shows a gap | It becomes a recommendation | It becomes next month's work, with an owner |
| What proves the work happened | The report itself | A dated link to each page, fix, listing and mention |
| What to ask to see | The report | The work log |
What the dashboard can and cannot show by month three
Strategi clients get a live dashboard instead of a monthly report, and what it answers changes across the 90 days. In month one it holds the baseline: each buyer question on each engine, who gets named instead and which pages AI crawlers read. Month two shows direction across weeks of daily readings, and month three sets the retest against the baseline beside first-party data. Strategi judges real movement at about 90 days, so month three is the first fair reading rather than a verdict.
First-party data carries more weight than any tracking estimate. According to eSEOspace, Google added Generative AI performance reports to Search Console on June 3, 2026, and Microsoft announced preview additions, including Citation Share, to AI visibility reporting in Bing Webmaster Tools on June 16, 2026. Search Console's AI reporting counts impressions, not clicks, so those numbers never appear as visits. Inquiries from AI assistants come from GA4, and pipeline value (AI inquiries multiplied by close rate and deal value) is shown as a floor, because some AI apps hide referrers.
Every tracked reading is a sample. LLM Pulse, a tracking vendor, calls prompt-tracking output "a repeatable sample of selected prompts, not a census of every AI answer". One AI answer is an observation, not a trend, and nobody can honestly say that one change produced one citation. The dashboard shows shipped work beside the trend around it and leaves the credit unclaimed.
Frequently asked questions
- What does a generative engine optimization agency actually do every month?
- A generative engine optimization agency should do five things every month: track a fixed list of buyer questions across AI engines, ship technical fixes, write new pages, improve existing pages and earn third-party mentions. Measurement is one of the five. The other four change something, and each should appear in a dated work log with links.
- Is AI search optimization just SEO with a new name?
- AI search optimization overlaps with SEO, most of all on Google. Google's May 2026 guidance calls optimizing for its AI features "still SEO", according to Business.com, which also notes that ChatGPT, Claude and Perplexity do not use Google's index. AI search optimization adds content carrying new information, clear answers near the top, planning around buyer questions and consistent facts across the web.
- What work does an AEO agency do after the first month?
- After the first month, an AEO (answer engine optimization) agency shifts most hours from setup to output: new answer pages and money pages, improvements to existing pages, off-site mentions where the plan includes them, and a month-three retest of the same buyer questions. Setup work such as crawler access and fact cleanup should shrink once month one is done.
- Should we keep our SEO agency and add an AI search agency on top?
- Usually not. Strategi treats SEO, AEO and generative engine optimization as one job, and two vendors publishing on the same site double the content and risk breaching Google's spam policies. An AI search optimization agency that does the full job replaces the SEO agency instead of sitting beside it, so one team owns every page.
- Can an AI search optimization agency guarantee ChatGPT will recommend us?
- No AI search optimization agency can honestly guarantee a mention or citation in ChatGPT or any other engine. GEO Catalyst, which sells software to agencies, says an agency can guarantee the measurement, implementation, retesting and reporting process, not a mention, citation, lead or revenue result. Strategi offers no guarantee on any plan.
- Is hiring an AI search optimization agency worth it?
- Hiring an AI search optimization agency is worth it when a company's buyers research with AI assistants before choosing a supplier. Google said in May 2026 that AI Mode had passed one billion monthly users worldwide, without a country breakdown. Where buyers do not use AI to research or buy, Strategi says no and usually points the business to outbound.
See your own baseline before you hire anyone
Every month of AI search work in this log starts from a baseline, and you can see yours before hiring anyone. Strategi's free dashboard shows the real answers ChatGPT, Gemini and Perplexity give to 10 of your buyer questions, who they name instead and where they look. Hold any agency's month-one plan against what it shows.