Agency LookoutThe call goes like this. The client opens Google Analytics, organic sessions are down 20% year over year, and they want to know what the agency has been doing. You check rankings: the same pages rank in the same positions. Impressions in Search Console are flat or up. Clicks are down.
This is now the most common SEO conversation in agencies, and it has a clear cause: a growing share of searches end in an AI answer (Google's AI Overviews and AI Mode, plus ChatGPT, Perplexity and the rest) without a click to any website. The page still ranks. It is just below an answer that already satisfied the searcher.
This guide covers how to see it in the data, what to track instead, and how to explain it so the client hears "the market changed and here's our plan" rather than "it's not our fault".
Step 1: Prove it with Search Console, not with a theory
Clients do not want a lecture on AI Overviews. They want evidence. Two reports make the case:
Impressions vs clicks, by query, over 16 months. Filter to queries where impressions held or grew while clicks fell. Those are the informational queries now being answered in place. If the client's drop concentrates in "how to", "what is", "best way to" queries, you have your explanation. If it is spread evenly across everything, including branded and transactional queries, something else is wrong (a site change, a penalty, a tracking break) and you should find that first.
Average position for those same queries. Stable position plus falling click-through rate is the signature of an AI answer sitting above the result. A falling position is an ordinary ranking problem and needs an ordinary fix.
Put the two side by side on one chart per query group. That picture is the whole argument.
Step 2: Separate the traffic that matters from the traffic that doesn't
Not all lost clicks were worth having. A how-to article that sent 2,000 visits a month who read one paragraph and left was producing impressions for the client's brand, not customers. The honest next question is: which queries that fell were producing leads or sales?
Pull conversions by landing page in GA4 against the pages that lost clicks. Usually you find:
- Informational pages lost the most clicks and were producing the fewest conversions.
- Commercial pages ("plumber near me", "pricing", product pages) held up, because AI answers send those searchers somewhere to act.
Say this plainly in the report. "Organic sessions fell 22%. Leads from organic fell 4%. The clicks we lost were mostly from informational searches that rarely converted." That reframing is accurate, and it changes the mood of the conversation.
Step 3: Measure visibility inside the AI answers
Rankings were never the goal; being the source a customer finds was. In 2026 that includes being cited inside AI answers. Track, per client, per month:
- Cited by AI answers: how many of the client's target queries produce an AI answer that cites the client's site.
- Business named by AI: for local and brand queries, whether the business itself is named in the answer.
- Share of answers vs competitors: who is cited instead, for the queries the client wants.
Several rank-tracking data providers now return AI Overview and AI Mode presence alongside classic positions, so this can be tracked the same way as rankings. Put the two numbers next to "keywords on page one" on the client's dashboard. "17 of your 40 target queries now cite you in the AI answer, up from 12" is a result the client can understand, and it is a number that keeps rising while classic clicks fall.
Step 4: Change what you report
The old headline metric, organic sessions, is now partly a measure of how many searches Google chose not to answer itself. Keep reporting it, but demote it. Lead with:
- Organic leads or revenue (the thing the client pays for).
- Impressions on target queries (demand is still there; this proves it).
- AI citations and business mentions (the new visibility).
- Clicks and sessions, with the AI-answer context explained once, in a footnote, not re-litigated every month.
Add one more line to the monthly report: pages gaining and losing clicks, with the query type labelled. Over a few months the client sees that the losses are in one category and the gains are in another, and stops asking.
Step 5: Adjust the work
What still moves the needle:
- Be the citable source. AI answers cite pages with clear, direct statements, specific numbers, and structure (headings that match the question, short definitions, tables). Rewrite the client's top informational pages so that the first paragraph answers the question outright. You are no longer writing to keep someone on the page; you are writing to be quoted.
- Own the commercial queries harder. They still send clicks. Local pages, service pages, pricing pages, comparison pages: this is where the SEO budget should move.
- Fix the fundamentals the AI systems check. Crawlability, schema (Organization, LocalBusiness, Product, FAQ), consistent business details across Google Business Profile and the site, fast pages. AI systems favour sources they can parse and trust.
- Make the brand searchable by name. When a customer sees the business named in an AI answer, the next thing they do is search the name. Branded search growth is a leading indicator; put it on the dashboard.
What to say on the call
A version that works:
"Your rankings are the same and your impressions are up, so demand for what you do hasn't dropped. What changed is that Google now answers a lot of informational searches on the results page, so fewer people click through. The clicks we lost were mostly on how-to pages that rarely turned into leads; leads from organic are down 4%, not 22%. We're tracking a new number from this month, how often your business is cited inside those AI answers, which is at 17 of your 40 target searches and rising. Our plan is to rewrite your top informational pages to be the source those answers quote, and to put more of the budget into the service pages, which still get the clicks and the leads."
That is a conversation about strategy, not about blame, and it ends with the client knowing what they are paying for.
Doing this for every client
The hard part is not the explanation; it is producing the evidence for 15 clients a month. Query-level impressions vs clicks, conversions by landing page, AI citation counts, branded search trends: four data sources, every account, every month. If you are assembling that by hand, it will not happen for the smaller accounts, and those are the ones that churn quietly.
Agency Lookout pulls Search Console, GA4 and AI visibility data into one place per client, shows pages gaining and losing clicks on the overview, tracks "cited by AI answers" and "business named by AI" next to rankings, and drops all of it into the monthly report. The explanation above becomes something the report already shows, rather than something you have to prove from scratch each time.