---
title: "AI Visibility for Agencies: The Timeline from Rank Trackers to Execution"
description: "Agencies moved from tracking AI snapshots to selling prompt dashboards. This timeline shows why monitoring is becoming a commodity and execution is the next test."
url: "https://groas.com/post/from-rank-trackers-to-execution-engines"
image: "https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/c969bbaa-6cec-4164-acbe-5e36928e2a10.png"
published: "2026-10-19T05:26:23.177Z"
modified: "2026-10-19T05:26:23.248Z"
---

October 19, 2026 · 10 min read

# AI Visibility for Agencies: The Timeline from Rank Trackers to Execution

[Alexander PerelmanHead Of Product @ groas](https://groas.com/author/alexander-perelman)[LinkedIn](https://www.linkedin.com/in/alexander-433793253/)

![Clay plumber stands in a flooding kitchen as a cheerful consultant on a ladder hands him a chart of the rising water, while a wrench sits unused.](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/c969bbaa-6cec-4164-acbe-5e36928e2a10.png)

In this article

1. [Late 2022: Perplexity makes cited answers visible, but not yet a retainer](#late-2022-perplexity-makes-cited-answers-visible-but-not-yet-a-retainer)
2. [May 2023: SGE gives rank trackers one more column](#may-2023-sge-gives-rank-trackers-one-more-column)
3. [May 2024: AI Overviews turn monitoring into a line item](#may-2024-ai-overviews-turn-monitoring-into-a-line-item)
4. [October 2024: ChatGPT Search makes prompt dashboards look indispensable](#october-2024-chatgpt-search-makes-prompt-dashboards-look-indispensable)
5. [2025: AI Mode and ads widen the gap between watching and acting](#2025-ai-mode-and-ads-widen-the-gap-between-watching-and-acting)
6. [Early 2026: Monitoring is easy to buy; execution is harder to fake](#early-2026-monitoring-is-easy-to-buy-execution-is-harder-to-fake)

Picture an $800-a-month AI visibility retainer: 120 prompts tracked, share-of-voice graphs, competitor mentions, and a report with a nice logo. Now ask what changed on the client’s site because of last month’s report. If the answer is nothing, you have the agency AI-visibility market in one awkward pause.

Since late 2022, the tooling has moved through three eras. First came rank trackers with AI snapshots bolted on. Then came prompt-monitoring dashboards. Now come platforms that execute the fixes. Agencies still packaging monitoring as the product are selling the middle era, **the part getting commoditized fastest**. The dates matter because each shift changed what an agency could reasonably charge for, and what work still landed back on its team.

You do not pick AEO software in a vacuum. You pick it inside a pricing model you have already sold. Two agencies can pay the same $489 a month for a tool and get opposite economics: one uses it to support a retainer; the other has to staff around everything the tool identifies but cannot fix. If you are weighing the [questions to ask before buying another dashboard](https://groas.com/post/reader-q-a-the-aeo-questions-agencies-ke), start with what each era actually delivered.

## Late 2022: Perplexity makes cited answers visible, but not yet a retainer

Perplexity was founded in August 2022 and [launched its answer engine on December 7, 2022](https://en.wikipedia.org/wiki/Perplexity_AI). By February 2023, it was doing about 2M visits. Answers could carry citations, which gave agencies something new to notice. It did not yet give them a useful service to sell: no familiar rank to report, no established prompt set to monitor, and little reason for a client to pay for screenshots of an answer engine.

That distinction matters. A citation is valuable to the business that earns it; a screenshot showing who got one is not automatically valuable to the business that did not. An agency trying to squeeze those screenshots into an SEO deck could describe the change, but the deck still could not tell anyone what to do next.

**Perplexity showed that AI answers could cite sources. Google would make those answers a routine client conversation.** Those are different milestones, and confusing them is how a new observation gets sold as a mature service.

## May 2023: SGE gives rank trackers one more column

Google announced Search Generative Experience at I/O in May 2023 as an opt-in Search Labs experiment. Its AI snapshot sat above the classic ten blue links. Rank trackers responded as they tend to respond to a new SERP feature: they bolted it on. If you already paid for rank tracking, you could get another column showing whether a snapshot appeared.

That column offered reassurance. Your rankings are still there; here is your AI snapshot presence, too. But **presence was not a task**. A checkbox saying Google generated an answer for 34 of your keywords did not tell a plumber or a SaaS founder what to rewrite on Tuesday morning. Nothing about the page changed because the tracker had spotted the box.

I used to argue that visibility was half the job. I was wrong. Visibility helps you decide where to work; it does not do the work. For an agency, that meant the add-on was useful as an alert, not much of a standalone retainer. The practical test was simple: could the client act on the new column, or did it merely give the account manager another thing to explain?

## May 2024: AI Overviews turn monitoring into a line item

Google rolled out [AI Overviews in the US starting May 14, 2024](https://www.techrepublic.com/article/google-io-2024-highlights/) as the renamed SGE, with broader availability within the week. AI answers were no longer confined to an opt-in experiment. They appeared in the results clients were already paying agencies to rank in. That made for an awkward SEO call: why was a paragraph answering the query sitting above our position-one result?

Agencies had an answer they could sell. For $200 to $500 a month on top of the retainer, a client could get AI Overview presence tracking, citation checks, and share of voice against two competitors. Monitoring became a product because the change was visible and clients wanted to know whether they were in the answer.

Then the product ran into the limit rank tracking had already taught us. A report saying you were cited in 11 of 60 AI Overviews, down from 14 last month, creates work for the agency and anxiety for the client. **It fixes nothing by itself.** Someone still has to decide whether the page needs rewriting, whether a technical block is in the way, or whether the answer is drawing from sources the client has not reached.

That is where the screenshot stopped carrying the retainer. The client wanted to appear in the answer, not keep paying to be told they were absent. Agencies that priced the observation as the whole service now had to supply the labor it exposed.

## October 2024: ChatGPT Search makes prompt dashboards look indispensable

OpenAI [launched ChatGPT Search on October 31, 2024](https://www.geekwire.com/2024/openais-new-chatgpt-search-engine-challenges-google-and-microsoft-and-further-upends-the-web/), starting with Plus and Team users plus a waitlist. Its inline citations looked a lot like Perplexity’s. That helped set the template for the second era: agencies no longer watched one Google box. They watched answers across four to six engines, each with a prompt set and a share-of-voice chart.

Dashboards made that spread manageable enough to sell. Vendors priced many of them by the prompt: $29 a month for a starter plan or $489 for a premium tier; $50 for 25 prompts here, $95 for 50 there; around $500 a month for roughly 250 prompts and a few workspaces at agency-tier tools. The numbers look tidy in a demo account. They get less tidy across a client roster.

Take ten clients with 80 prompts each. You pay the vendor for coverage, then pay staff to explain why prompt 47 moved while prompt 48 did not. Add an engine or expand a prompt set, and you have more observations to sort before anyone touches a page. The dashboard solved the _where do we look?_ problem while making the _who will act?_ problem harder to ignore.

**Tracking more engines did not automatically create more value.** It could just create more screenshots to explain. If an agency could not connect those observations to changes it made, the prompt count was a cost center dressed as a deliverable.

## 2025: AI Mode and ads widen the gap between watching and acting

Google [started testing AI Mode on March 5, 2025](https://thatware.co/google-to-appear-on-google-ai-mode/) as an early experiment for complex questions, expanded it to all US users in May, then pushed it to more than 35 new languages and 40-plus territories by autumn for over 200 total. For agencies, the important change was architectural. AI Mode was not another box above results. It was a separate conversational tab with its own citations and follow-ups. A page cited in an AI Overview was not automatically cited in AI Mode, so a prompt set built for one surface could miss the other.

In parallel, the answer became paid space. Google [began testing ads in AI Overviews in May 2024](https://seranking.com/blog/ai-overviews/), rolled out the first mobile-US placements in October and November 2024 labeled Sponsored, then expanded to desktop in May 2025 and wider global coverage later that year. Once Google could sell clicks inside the answer, an agency watching that answer without a way to influence paid or organic presence had an increasingly narrow service.

This is where the middle era began to look expensive. The practitioner had more surfaces to check, more prompts to maintain, and more reports to reconcile. The client still asked the same question: what are you changing? **The answer could no longer be another prompt count.** The work had to reach the page, the technical setup, the citation, or the paid campaign. Otherwise the dashboard was a better map of a problem the agency still had to solve by hand.

## Early 2026: Monitoring is easy to buy; execution is harder to fake

By early 2026, agency tool comparisons read like a price list for variations on monitoring: Scrunch Agency at $500 a month for roughly 250 prompts and three workspaces, Trakkr Scale near $500 for ten brands with portals, Peec Pro from $245 plus credits, and Profound on custom enterprise [pricing that assumes you have staff to act on what it finds](https://citetrackai.com/blog/best-ai-visibility-tools-for-agencies/). They report. The work the reports identify still largely needs doing. White-label portals and per-client reporting appear alongside prompt allowances and predictable per-client cost. When the logo on the PDF becomes a selling point, the PDF is not much of a moat.

![Printed AI visibility reports beside a laptop showing published fixes](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/a2d2eec1-4b61-4469-a04a-4b801789c7ba.png)

The economics get uncomfortable across a book of business. Say you run 15 clients at 80 prompts each. On capped plans, you ration prompts, rotate which clients get tracked, or pay overages that eat into the margin you added for reporting. The [math on prompt caps versus per-domain pricing](https://groas.com/post/flat-rate-aeo-pricing-per-domain-9c1ae) looks different when you run it for every client instead of one demo account. Monitoring scales cost. Execution is the work a client actually bought the retainer for.

The timeline makes a useful buying filter:

- **Era one, rank-tracker add-ons:** an AI snapshot appeared for 34 keywords. The extra column named the surface but assigned no fix.
- **Era two, prompt dashboards:** share of voice moved from 12% to 9%. The chart showed the change but did not publish a page, fix schema, or pursue a citation.
- **Era three, execution:** monitoring points to work that gets done. Judge the platform by the changes it ships, not the number of prompts it can display.

![Whiteboard timeline from AI search tracking charts to published fixes](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/2dd85333-15d5-4d3a-bd87-9acfe6e15ec1.png)

### The test: show me what shipped

When I look at a platform now, I ask one question: **what shipped last week without my team touching it?** If the answer is a PDF, the product belongs to the monitoring era. Reporting that share of voice fell from 14% to 11% is useful only if someone can change why: write pages for the buyer questions the engines draw on, remove technical blocks that keep those pages out of reach, and work on the citations the answers use.

That is the distinction [groas makes for agencies](https://groas.com/for-agencies): execution under your brand, with bids, budgets, content, technical fixes, and citations handled around the clock; actions logged with their reasoning; and weekly reports ready to forward with your logo. The report still matters. It tells the client what happened. It is no longer being asked to stand in for the work.

![White-label client report beside a checklist of published fixes](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/cf6266d9-0a37-442f-a5a3-528ad2697942.png)

### The cost test: can you look closer without paying more?

Prompt caps tax closer inspection. If following 80 buyer questions across ChatGPT, AI Overviews, and Perplexity forces a tier jump for each client, the temptation is to skip long-tail follow-ups. Per-domain pricing changes that incentive: you pay for the client rather than rationing which questions deserve a look. Across 15 clients, a capped monitoring bill can climb as coverage expands; a per-domain model keeps that part of the cost predictable.

The point is not that every prompt deserves a report. It is that **a vendor should not make you look less closely to protect your margin**. Otherwise your service gets thinner as your understanding of the client’s questions gets better. That is a strange way to price diligence.

### The operations test: does the fix leave the queue?

The last bottleneck is implementation. Era-two tools assumed your developers or the client’s developers would publish whatever the dashboard found. When that work sat in a queue, the next report measured the same problem again. A portal could make the backlog look professional. It could not clear it.

For a multi-client agency, the practical test is whether each client’s ad accounts and website can be connected without turning every fix into a development project, and whether content, technical changes, and citations can be executed behind the agency’s brand and logged. I no longer judge AEO software by how many engines it tracks. I judge it by how many client domains it can work on without asking my team to become the implementation layer.

That is where the line points next: monitoring stays necessary, but it stops being the thing worth a retainer on its own. Agencies still selling screenshots are standing on the wrong side of it.

## Frequently Asked Questions

### How has AI visibility tooling for agencies changed since 2022?

The tooling moved through three eras since late 2022: rank trackers with AI snapshots bolted on, then prompt-monitoring dashboards, and now platforms that execute the fixes. Agencies still packaging monitoring as the product are selling the middle era, the part getting commoditized fastest.

### Why did Perplexity's answer engine not immediately become a service agencies could sell?

Perplexity launched its answer engine on December 7, 2022, and answers could carry citations, which gave agencies something new to notice. But it offered no familiar rank to report, no established prompt set to monitor, and little reason for a client to pay for screenshots. A citation is valuable to the business that earns it; a screenshot showing who got one is not automatically valuable to the business that did not.

### What did rank trackers do when Google's SGE launched in May 2023?

Rank trackers bolted the AI snapshot on as an extra column showing whether a snapshot appeared, alongside your existing rankings. That column offered reassurance but presence was not a task: knowing Google generated an answer for 34 of your keywords did not tell anyone what to rewrite. The add-on was useful as an alert, not much of a standalone retainer.

### How did AI Overviews turn AI monitoring into a paid product for agencies?

Google rolled out AI Overviews in the US starting May 14, 2024, and AI answers appeared in the results clients were already paying agencies to rank in. Agencies could then sell AI Overview presence tracking, citation checks, and share of voice against two competitors for $200 to $500 a month on top of the retainer. But a report saying you were cited in 11 of 60 AI Overviews fixes nothing by itself; someone still has to decide what to rewrite or which technical block is in the way.

### Did tracking more AI engines create more value for agencies after ChatGPT Search launched?

Not automatically. After OpenAI launched ChatGPT Search on October 31, 2024, agencies watched answers across four to six engines, with prompt-based vendor pricing like $29 a month for a starter plan or $489 for a premium tier. Across ten clients with 80 prompts each, you pay for coverage and then pay staff to explain why prompt 47 moved. If observations are not connected to changes, the prompt count is a cost center dressed as a deliverable.

### Why did Google AI Mode and ads in AI answers change how agencies see monitoring?

AI Mode, tested from March 5, 2025 and later expanded to over 200 languages and territories, was a separate conversational tab with its own citations, so a page cited in an AI Overview was not automatically cited in AI Mode. In parallel, Google began testing ads in AI Overviews and expanded Sponsored placements from mobile US to desktop and globally by 2025. Watching the answer without a way to influence paid or organic presence left agencies with an increasingly narrow service.

### How should an agency judge an AI visibility platform?

Ask what shipped last week without your team touching it. If the answer is a PDF, the product belongs to the monitoring era. Judge the platform by the changes it ships, such as writing pages for buyer questions, removing technical blocks, and working on citations, not by the number of prompts it can display.

### Why do prompt caps hurt agencies, and what pricing model avoids that?

Prompt caps tax closer inspection: if following 80 buyer questions across ChatGPT, AI Overviews, and Perplexity forces a tier jump for each client, the temptation is to skip long-tail follow-ups. Across 15 clients, a capped monitoring bill can climb as coverage expands. Per-domain pricing changes the incentive: you pay for the client rather than rationing which questions deserve a look, keeping that part of the cost predictable.

## Pay For Results, Not For Hours

Businesses buy the outcome, agencies resell it, and groas answers for it either way.

[See If You Qualify](https://groas.typeform.com/to/xC1bQNUT)

## Structured data

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://graph.groas.com/entity/groas","name":"groas","url":"https://groas.com/","logo":{"@type":"ImageObject","url":"https://groas.com/icon-512.png","width":512,"height":512},"sameAs":["https://www.linkedin.com/company/groas/"]},{"@type":"WebSite","@id":"https://groas.com/#website","name":"groas","url":"https://groas.com/","inLanguage":"en-US","publisher":{"@id":"https://graph.groas.com/entity/groas"}},{"@type":"WebPage","@id":"https://groas.com/post/from-rank-trackers-to-execution-engines#webpage","url":"https://groas.com/post/from-rank-trackers-to-execution-engines","name":"AI Visibility for Agencies: The Timeline from Rank Trackers to Execution","description":"Agencies moved from tracking AI snapshots to selling prompt dashboards. This timeline shows why monitoring is becoming a commodity and execution is the next test.","inLanguage":"en-US","isPartOf":{"@id":"https://groas.com/#website"},"about":{"@id":"https://graph.groas.com/entity/groas"},"primaryImageOfPage":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/c969bbaa-6cec-4164-acbe-5e36928e2a10.png"},"breadcrumb":{"@id":"https://groas.com/post/from-rank-trackers-to-execution-engines#breadcrumb"},"dateModified":"2026-10-19T05:26:23.248Z","mainEntity":{"@id":"https://groas.com/post/from-rank-trackers-to-execution-engines#article"}},{"@type":"BreadcrumbList","@id":"https://groas.com/post/from-rank-trackers-to-execution-engines#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://groas.com/"},{"@type":"ListItem","position":2,"name":"Blog","item":"https://groas.com/blog"},{"@type":"ListItem","position":3,"name":"AI Visibility for Agencies: The Timeline from Rank Trackers to Execution","item":"https://groas.com/post/from-rank-trackers-to-execution-engines"}]},{"@type":"BlogPosting","@id":"https://groas.com/post/from-rank-trackers-to-execution-engines#article","url":"https://groas.com/post/from-rank-trackers-to-execution-engines","mainEntityOfPage":{"@id":"https://groas.com/post/from-rank-trackers-to-execution-engines#webpage"},"isPartOf":{"@id":"https://groas.com/blog#blog"},"headline":"AI Visibility for Agencies: The Timeline from Rank Trackers to Execution","description":"Agencies moved from tracking AI snapshots to selling prompt dashboards. This timeline shows why monitoring is becoming a commodity and execution is the next test.","datePublished":"2026-10-19T05:26:23.177Z","dateModified":"2026-10-19T05:26:23.248Z","image":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/c969bbaa-6cec-4164-acbe-5e36928e2a10.png"},"author":{"@id":"https://groas.com/author/alexander-perelman#person"},"publisher":{"@id":"https://graph.groas.com/entity/groas"},"keywords":"agencies, prompt, rank, execution, client, rank trackers, answer, monitoring","wordCount":2036,"inLanguage":"en-US"},{"@type":"Person","@id":"https://groas.com/author/alexander-perelman#person","name":"Alexander Perelman","jobTitle":"Head Of Product @ groas","description":"My background includes Goldman Sachs and Stanford Computer Science.","image":"https://groas.com/media/blog/4f17cac3a81acc224cc1b2cfab8f94bbc295a3089d13e8421125771ef8d4064f.jpg","worksFor":{"@id":"https://graph.groas.com/entity/groas"},"sameAs":["https://www.linkedin.com/in/alexander-433793253/"]},{"@type":"FAQPage","@id":"https://groas.com/post/from-rank-trackers-to-execution-engines#faq","mainEntity":[{"@type":"Question","name":"How has AI visibility tooling for agencies changed since 2022?","acceptedAnswer":{"@type":"Answer","text":"The tooling moved through three eras since late 2022: rank trackers with AI snapshots bolted on, then prompt-monitoring dashboards, and now platforms that execute the fixes. Agencies still packaging monitoring as the product are selling the middle era, the part getting commoditized fastest."}},{"@type":"Question","name":"Why did Perplexity's answer engine not immediately become a service agencies could sell?","acceptedAnswer":{"@type":"Answer","text":"Perplexity launched its answer engine on December 7, 2022, and answers could carry citations, which gave agencies something new to notice. But it offered no familiar rank to report, no established prompt set to monitor, and little reason for a client to pay for screenshots. A citation is valuable to the business that earns it; a screenshot showing who got one is not automatically valuable to the business that did not."}},{"@type":"Question","name":"What did rank trackers do when Google's SGE launched in May 2023?","acceptedAnswer":{"@type":"Answer","text":"Rank trackers bolted the AI snapshot on as an extra column showing whether a snapshot appeared, alongside your existing rankings. That column offered reassurance but presence was not a task: knowing Google generated an answer for 34 of your keywords did not tell anyone what to rewrite. The add-on was useful as an alert, not much of a standalone retainer."}},{"@type":"Question","name":"How did AI Overviews turn AI monitoring into a paid product for agencies?","acceptedAnswer":{"@type":"Answer","text":"Google rolled out AI Overviews in the US starting May 14, 2024, and AI answers appeared in the results clients were already paying agencies to rank in. Agencies could then sell AI Overview presence tracking, citation checks, and share of voice against two competitors for $200 to $500 a month on top of the retainer. But a report saying you were cited in 11 of 60 AI Overviews fixes nothing by itself; someone still has to decide what to rewrite or which technical block is in the way."}},{"@type":"Question","name":"Did tracking more AI engines create more value for agencies after ChatGPT Search launched?","acceptedAnswer":{"@type":"Answer","text":"Not automatically. After OpenAI launched ChatGPT Search on October 31, 2024, agencies watched answers across four to six engines, with prompt-based vendor pricing like $29 a month for a starter plan or $489 for a premium tier. Across ten clients with 80 prompts each, you pay for coverage and then pay staff to explain why prompt 47 moved. If observations are not connected to changes, the prompt count is a cost center dressed as a deliverable."}},{"@type":"Question","name":"Why did Google AI Mode and ads in AI answers change how agencies see monitoring?","acceptedAnswer":{"@type":"Answer","text":"AI Mode, tested from March 5, 2025 and later expanded to over 200 languages and territories, was a separate conversational tab with its own citations, so a page cited in an AI Overview was not automatically cited in AI Mode. In parallel, Google began testing ads in AI Overviews and expanded Sponsored placements from mobile US to desktop and globally by 2025. Watching the answer without a way to influence paid or organic presence left agencies with an increasingly narrow service."}},{"@type":"Question","name":"How should an agency judge an AI visibility platform?","acceptedAnswer":{"@type":"Answer","text":"Ask what shipped last week without your team touching it. If the answer is a PDF, the product belongs to the monitoring era. Judge the platform by the changes it ships, such as writing pages for buyer questions, removing technical blocks, and working on citations, not by the number of prompts it can display."}},{"@type":"Question","name":"Why do prompt caps hurt agencies, and what pricing model avoids that?","acceptedAnswer":{"@type":"Answer","text":"Prompt caps tax closer inspection: if following 80 buyer questions across ChatGPT, AI Overviews, and Perplexity forces a tier jump for each client, the temptation is to skip long-tail follow-ups. Across 15 clients, a capped monitoring bill can climb as coverage expands. Per-domain pricing changes the incentive: you pay for the client rather than rationing which questions deserve a look, keeping that part of the cost predictable."}}]}]}
```