AI SEO Case Study: How I Grew Organic Clicks 4.8x in 9 Months
Between January and September 2026, my own site went from 1,322 Google clicks a month to 6,412, a 4.8x increase, and from 32 visits a month sent by AI assistants such as ChatGPT, Perplexity, Claude and Copilot to 3,599. Google impressions rose from 209,381 to 831,210 a month. This is my SEO and AI visibility case study of lilachbullock.com: what I changed, in what order, and what it means for your business.
Every number on this page comes from my own Google Search Console and Google Analytics 4, pulled on 7 October 2026, using full calendar months. I'm Lilach Bullock, and I test everything on my own site before I recommend it to anyone else. This is the result.
The results at a glance
| Measure | January 2026 | September 2026 | Change |
|---|---|---|---|
| Google clicks per month | 1,322 | 6,412 | 4.8x |
| Google impressions per month | 209,381 | 831,210 | 4.0x |
| Average position in Google | 27.2 | 16.5 | 10.7 places higher |
| Visits from AI assistants per month | 32 | 3,599 | over 112x |
| Of which from ChatGPT | 29 | 3,267 | over 112x |
| New referring domains in the month | 40 | 304 | 7.6x |
Month by month
The work started in earnest in May. Google clicks have gone up every single month since.
| Month (2026) | Google clicks | Clicks chart | Impressions | Avg position | AI assistant visits |
|---|---|---|---|---|---|
| January | 1,322 | 209,381 | 27.2 | 32 | |
| February | 937 | 142,439 | 21.4 | 42 | |
| March | 963 | 132,683 | 26.1 | 61 | |
| April | 1,011 | 131,239 | 21.7 | 22 | |
| May | 1,226 | 291,973 | 17.4 | 32 | |
| June | 2,210 | 444,725 | 16.6 | 85 | |
| July | 3,836 | 243,818 | 25.7 | 151 | |
| August | 4,793 | 503,320 | 25.6 | 982 | |
| September | 6,412 | 831,210 | 16.5 | 3,599 |
Where I started
In January my site had over 7,000 posts built up over many years, 1,322 Google clicks a month and an average position of 27.2, which is page three. AI assistants sent it 32 visits in the whole month. New referring domains were arriving at between 26 and 40 a month. A lot of good content, and most of it working far below its potential. That's the position I see most established business sites in.
What I did
1. Technical fixes first
Before pushing anything, I cleared what was stopping Google from crawling and trusting the site. In May I removed more than 700 thin tag archive pages from the index, fixed duplicate H1 headings on key pages, repaired broken links and pointed redirected links straight at their final pages, wrote alt text for more than 7,000 images and rewrote hundreds of titles and meta descriptions that were out of date or cut off in the results. None of it's glamorous. All of it matters, because Google only has so much time for each site.
2. Topic hubs and internal linking
From June I grouped related posts into topic hubs. Each hub links to every page in its topic and each page links back, with anchor text that describes the page it points to. Every target page has a ceiling on how many new links it gets, so nothing looks spammed. Internal links are how Google learns what a page is about and which pages matter most, and they're the one ranking signal you fully control.
3. Page-one pushes
The quickest wins sit on pages that already rank in positions 4 to 15 for searches with plenty of impressions. For those pages I matched the title to the words people search with, added the answer the page was missing and pointed relevant internal links at them. One rule I never break: don't retitle a page that's already in the top three. I've seen a title change for "better click-through" knock a page off the top spot within days.
4. Answer-first content
In September I added answer-first openings to 244 pages: two or three sentences at the top that answer the question in the title directly, before anything else. It's what readers want, it's what Google uses for featured snippets and it's the passage AI assistants are most likely to quote. One answer block per page is enough. Stacking summaries, key-facts boxes and short versions on top of each other looks like a page built for machines.
5. AI citation work
To be named in AI answers, the AI has to understand who you are. I published llms.txt and llms-full.txt files that describe the site and its best pages, a fact sheet page about me and my work, and structured data that ties my articles to me as the author. I also watched which pages AI assistants were already sending people to and built more pages on those topics. ChatGPT visits went from 29 in January to 3,267 in September.
6. Indexing
Google can't rank a page it hasn't crawled, and a change doesn't count until it's been seen. Every new or updated page is submitted to Bing and IndexNow straight away and put at the front of the queue for Google's recrawl, and I check what Google has and hasn't indexed. I measure results after the recrawl, never just after the edit.
7. A content engine steered by data
New content goes out every day, but the topics aren't guesses. They come from the searches I already get impressions for without a page to match, the questions people ask AI assistants and topics close to pages that are already winning. Every topic has to pass a relevance check before it's written. Much of the daily work, from indexing to internal links, runs on AI agents I built, so the method keeps working while I work with clients.
8. Authority
Alongside the on-site work I published pages other sites want to cite, such as statistics round-ups, a free website directory and original studies. In a third-party backlink index, new referring domains went from 40 in January to 304 in September.
The timeline
| When | What I did | Google clicks that month |
|---|---|---|
| January to April | The baseline: the site before the work started | 1,322 to 1,011 |
| May | Technical clean-up: thin archives out of the index, duplicate headings, broken and redirected links, image alt text, titles and descriptions | 1,226 |
| June | Topic hubs and internal linking; the content engine pointed at the AI topics people search for | 2,210 |
| July | Titles trimmed to fit the results page on around 800 pages, missing descriptions written; daily indexing pushes start | 3,836 |
| August | llms.txt rebuilt; extra summary blocks cleaned up so each page has one answer; AI visits reach 982 | 4,793 |
| September | Answer-first openings on 244 pages; new pages for searches with no page; hub wiring; 304 new referring domains | 6,412 |
What this means for your business
- Your next customers are probably already seeing you on page two. Most established sites have pages ranking just off page one. Pushing those is faster than starting new pages from scratch.
- AI assistants are a real source of visitors now, and it's growing fast. My AI traffic grew more than 112 times in nine months. The businesses that AI tools understand and trust are the ones they recommend.
- The same work wins on Google and in AI answers. Clean indexing, clear answers, strong internal links and a named expert behind the page help with both.
- Order matters. Fix first, push the pages that are close, then publish. Doing it the other way round wastes months.
- It compounds. Each month built on the one before. Nothing here was a one-off trick.
How I measured this
Google clicks, impressions and average position are whole-site totals from Google Search Console for web search, by calendar month. AI assistant visits are Google Analytics 4 sessions where the source is an AI assistant, including ChatGPT, Perplexity, Claude, Copilot, Gemini and Meta AI. Referring domains come from a third-party backlink index. All figures were pulled on 7 October 2026.
Frequently asked questions
How long did it take to see results?
The first clear lift came in June, about a month after the technical clean-up in May: Google clicks went from 1,226 to 2,210. Clicks have grown every month since, reaching 6,412 in September. Changes only count once Google has recrawled the page, so expect weeks, not days.
Is traffic from AI assistants real traffic?
Yes. These are visits recorded in Google Analytics 4 where the source is an AI assistant such as chatgpt.com, perplexity.ai, claude.ai, copilot.com or gemini.google.com. Each one is a person who clicked a link to my site inside an AI answer.
Does SEO still matter now that people use AI search?
More than ever, but the job has changed. AI assistants lean on pages that search engines already trust, so the same work that lifts your Google rankings also makes you easier to cite: clean indexing, clear answers, strong internal links and a named expert behind the page.
Can this work for a small business website?
Yes. A smaller site has fewer pages, but the order of work is the same: fix what stops Google crawling and trusting the site, push the pages that are already close to page one, answer the questions your buyers ask, and make it easy for AI tools to understand who you are and what you do.
What did you do first?
The technical fixes. There's no point pushing pages or publishing more while Google is wasting its time on thin archive pages, duplicate headings and broken links. Clean foundations first, then growth work.
Want the same for your site?
I do this work for business owners directly. If you want more traffic from Google and to be named in AI answers, start with my AI SEO services, or see how my generative engine optimization services and answer engine optimization services work. Not sure where you stand? My AI visibility audit shows what AI tools say about you today, and the free AI Visibility Checker gives you a quick first read.
Ready to talk? Get in touch and tell me about your site. I'll tell you plainly where your biggest wins are.