Google traffic has collapsed since AI Overviews showed up. Should I stop doing SEO and move everything to ads?

Organic clicks are genuinely down and it is not your imagination. But the advice going around right now — rebuild everything for “GEO”, or give up and pour it all into Meta — is mostly being sold by people who benefit from you believing it. Here is what the numbers actually say and what I would do with a small Indian budget.

Quick answer if you’re skimming

Clicks on the #1 organic result fall about 58% when an AI Overview sits above it. That part is real.
The traffic is not moving to ChatGPT. In August 2026 Google was 88.55% of search referrals; ChatGPT was 0.531%. All AI combined stayed under 0.9%.
So Google is taking clicks away and AI is not yet giving them back. Anyone selling GEO as a like-for-like replacement is overselling.
Moving the budget to ads is not free: 18% GST on spend, plus 10–20% agency fee on top.
Informational top-of-funnel content is what died. Pages that answer a buying decision still get clicked.

How much traffic am I actually losing to AI Overviews?

Clicks to the top organic result drop roughly 58% when an AI Overview is present (Ahrefs).
With an AI Overview on the page, 8% of users click a traditional result, against 15% without one (Pew Research Center).
Around 83% of AI Overview searches end with no click at all (Similarweb).
AI Overview coverage is not a one-way ramp. It peaked near 25% of queries in mid-2025 and fell to about 7% by late 2025. Google tunes this constantly.
Practical read: do not build a plan around one month’s snapshot. The dial moves.

Is the traffic going to ChatGPT instead?

No, and this is the number nobody quotes. August 2026 search referrals: Google 88.55%, ChatGPT 0.531%, Claude 0.30%, all AI combined under 0.9%.
ChatGPT’s referral share is bouncing, not climbing: 377% in June, 0.913% in July, 0.531% in August 2026.
Ahrefs, a company that sells AI-search tooling, gets about 5% of its own traffic from AI search.
India is ahead on adoption — a July 2026 YouGov study put 89% of urban Indian internet users on AI chatbots for search — but only 27% treat AI as their primary source, and 36% use it after a normal search.
So the clicks are disappearing, not relocating. That is an uncomfortable answer but it is the honest one.

Then why is everyone telling me I need GEO or AEO right now?

Because it is the easiest thing to sell in a year when everyone’s traffic is down and nobody can explain it.
There is a real version of it. Traffic that does arrive from AI converts much better — Semrush puts an AI-search visitor at roughly 4x the value of an average organic visitor.
But the volume is tiny, and 7% of ChatGPT queries are informational with about 0.1% transactional. Do not model it as bottom-of-funnel revenue.
One genuinely useful finding: ChatGPT cites pages ranking at position 21 or lower about 90% of the time. Being in the top three is not what gets you cited.
Another: roughly 50% of links in ChatGPT answers point to business and service sites, not publishers. Product and home pages over-perform in AI search relative to organic.
So treat GEO as cheap insurance you spend a few hours on, not a budget line that replaces SEO.

Did the August 2026 spam update hit me rather than AI Overviews?

Google ran a spam update from 18 to 21 August 2026 targeting scaled content abuse. If your drop is dated in that window, this is the likelier cause.
Documented casualties: a YMYL site losing rankings across 200,000+ queries, a programmatic affiliate site losing 14,000, and a 1.5 million-URL site that was roughly 85% programmatic hit site-wide.
The common factor was scale without added value, not the use of AI tools. AI-assisted pages that contained original information were not the target.
If you published a few hundred near-identical location or “best X in Y” pages this year, that is your answer.

How do I tell which one it is?

Pull Search Console, set the date range either side of 18 August 2026, and compare impressions against clicks.
Impressions flat, clicks down = you are still ranking, people are not clicking. That is AI Overviews or a SERP layout change.
Impressions down too = you lost rankings. That is an algorithm hit, and the August spam update is the first suspect.
Then split by page type. If the bleeding is all in informational blog posts and your product pages are steady, that is the normal 2026 pattern and not an emergency.
Do this before you spend a rupee on a fix. Half the “we need to rebuild the site” panics dissolve at this step.

Should I just move the budget to Google and Meta ads?

Budget in gross, not net. 18% GST applies on ad spend, so a ₹30,000 budget is a ₹35,400 outlay.
Good news that a lot of people are still getting wrong: the 6% equalisation levy on foreign digital advertising was abolished from 1 April 2025. If anyone is still quoting a 6% surcharge on your Google or Meta invoice, that is out of date.
Small budgets stall. Below roughly ₹30,000/month in spend, campaigns spend most of their life in the learning phase and never stabilise.
Indian CPCs vary close to tenfold by vertical. Local services sit at the low end; finance, insurance and legal at the top. Get your own number from Keyword Planner before you model anything.
The honest trade: ads buy you traffic you stop receiving the day you stop paying. SEO buys you an asset that is currently depreciating. Neither is a clean win in 2026.

What does an agency actually cost in India, and is it worth it?

Typical Indian SMB retainers run ₹8,000–₹50,000/month, or 10–20% of ad spend. Advanced scopes go well past ₹80,000.
Fees usually add 20–40% on top of spend. At ₹30,000 of ads you are really committing closer to ₹45,000 all-in with GST.
For reference, GrowthLife publishes its pricing openly and its SEO/GEO tier starts at $700/month, roughly ₹62,000. I am quoting that not as a pitch but because most Indian agencies will not give you a band until the third call, and you should know where the market sits before you start.
At the ₹30,000–₹50,000 total level you are usually better off with one good freelancer than a junior at an agency.
The question that filters agencies fast: ask for a case study with actual numbers attached. A lot of “real results” pages turn out to be qualitative only.

So what would actually work right now?

Stop writing top-of-funnel explainers. That is the traffic AI Overviews absorbed and it is not coming back.
Write the pages a person reads with a card in their hand: comparisons, pricing, “is X worth it”, integration and compatibility pages. Those still get clicked because the answer box cannot close the decision.
Get your prices, specs and FAQs into clean text on the page rather than inside images or JS. That is most of what “GEO” actually means in practice.
Build the channels nobody can deprecate: email, WhatsApp, a repeat-purchase habit. Indian D2C brands running WhatsApp-led retention consistently report materially better repeat rates than email alone.
Measure blended, not per-channel. Judging Meta on Meta-attributed ROAS alone is the single most common reason founders think it stopped working when it did not.

Where the money goes, honestly

Option	What it really costs	What you get	Honest verdict
Google / Meta ads	Spend + 18% GST + 10–20% mgmt	Traffic today	Works, stops the day you stop paying
Classic SEO (blog volume)	Writer + time	Shrinking returns	The part that actually died
Decision-stage pages	Low, mostly effort	Fewer visits, better ones	Best return in 2026
GEO / AEO retainer	₹25k–₹60k+/mo	Under 1% of referrals today	Do the free 80%, skip the retainer
Email / WhatsApp	Low and fixed	Owned, unkillable	Most underrated line here

What I’d actually do

Spend an hour in Search Console splitting impressions from clicks either side of 18 August 2026. You cannot fix a drop you have not diagnosed, and the fix is completely different in each case.
Move content effort from “what is X” to “should I buy X, and which one”. Keep the same budget, change the target.
Do not sign a GEO retainer this quarter. Put clean text on your key pages yourself, then revisit in six months when the referral share is worth paying for.