The Botfusions GEO Process: What We Actually Do

Correction (6 October 2026): Our measurement provider withdrew two data sets used in this post. The 5 October 2026 scan was only partially measured: the provider ran out of credit during the scan; ChatGPT answered all 300 questions, while Gemini, Perplexity, Claude and Grok answered only 42–44. The Gemini and Google AI Overview zeros in the 15 September 2026 scan were also a measurement error. Those numbers in this post are invalid. The 15 September ChatGPT measurement (249 answers, 11.2%, ±3.9 points) stands. We will update this post when a complete new scan is available.

Why "Ten Blue Links" Is No Longer the Whole Game

For two decades, SEO meant ranking on a search results page. Since 2024, that page increasingly includes a synthesized answer from ChatGPT, Gemini, Claude, or Google's AI Overview — generated before the user ever sees a list of links. To be visible in that answer, your brand has to be found, judged reliable, and cited. This is what Generative Engine Optimization (GEO) is for.

This page used to describe a "5-phase proprietary framework" with invented acronyms and a fabricated case study. We took it down. What follows is what we actually do.

How We Actually Run the Process

The work is done by people using measurement tools, not by autonomous agents.

1. Measure first

Every engagement opens with a scan: commercial-intent queries run across the AI engines we cover, recording for each response whether the brand was mentioned, its position, sentiment and citation share. The engine set and query count are written down with every scan, and the margin of error is published alongside the result. Our own numbers, including the misses, are public: AI Visibility Benchmark Report 2026.

2. Fix the technical layer

Schema (JSON-LD), llms.txt, entity grounding, crawlability and page-level citability. This is engineering work on your site, not a dashboard handed over with a to-do list.

3. Re-measure in the same frame

The second scan repeats the first one as closely as possible: same query set, same conditions. The measurement platform can change the engine set from scan to scan; when it does, we do not compare the two scans as a whole, only the engines that ran in both, one by one. A change smaller than the margin of error is not reported as a result.

What one round looks like: botfusions.com, September 15 – October 5, 2026

Here is how we ran the process on our own site, with dates. Every figure comes from our own scans.

September 15, 2026 · measure. The scan ran ChatGPT, Gemini and Google AI Overview. We were mentioned in 11.2% of 249 ChatGPT responses (±3.9 points). We were not mentioned at all on Gemini (379 responses) or Google AI Overview (372 responses).

September 16 – October 5, 2026 · fix. The main changes we made to the site in this window:

• September 16: removed invented internal source names and fake reviews. Added an unsourced-figure check to the build.
• September 21: took down 12 blog posts that contained invented method names and fake case studies, and rewrote this page. Shipped the internal linking table to code.
• September 22: fixed a fake freshness signal that wrote "updated today" on every build on two pages, with no content change.
• September 24: restructured the About page for AI search.
• September 26: fixed the findings of a Semrush site audit: invalid schema properties and wrong hreflang tags.
• September 28–29: moved Turkish service pages to Turkish URLs with 301 redirects. Made robots.txt disallow rules apply to AI bots too.
• September 30: added a founder author box, author data on guides and external sources.
• October 5: removed an "8 engines" phrase that overstated our measurement coverage.

October 5, 2026 · re-measure. The engine set changed in this scan: ChatGPT, Gemini, Perplexity, Claude and Grok ran; Google AI Overview did not. So we did not compare the two scans as a whole. We read the engines that ran in both, one by one:

• ChatGPT: 11.2% → 16.0% (300 responses, ±4.1). The difference is 4.8 points, the combined margin is ±5.7 points. We do not count it as a change.
• Gemini: 0% → 11.9% (5 of 42 responses, ±9.9). The difference exceeds the combined margin, but the sample is small. We do not count it as a result until the next scan confirms it.
• Google AI Overview: not measured. We have nothing new to say about it.

We cannot say which of the changes above caused these differences. We changed several things in the same window and the engine set changed too. What we took from this round is about method.

What we learned from this round

1. The engine set is not under our control. So we write the engines measured that day at the top of every report and report an unmeasured engine as "not measured", not 0%.
2. A small sample makes you happy too fast. The Gemini increase rests on 42 responses. We do not present it to clients as a win before it is confirmed.
3. Branded queries inflate the rate. Most queries in the citation sample were of the form "Botfusions or X". We are asking for the unbranded rate separately.
4. The language of the cited page matters. In the October 5 citation sample, 13 of 25 citations went to our English pages. Why the Turkish pages were not chosen is the next task.

We explain how we build the report in our AI visibility report post and the multi-engine monitoring workflow in our AI visibility monitoring post.

Frequently Asked Questions

How is GEO different from traditional SEO?

Traditional SEO ranks you on a results page. GEO is about being the synthesized answer an AI agent gives — which depends on your content being retrievable, well-structured, and judged reliable when the model looks for a source.

Is GEO a replacement for SEO?

No. Google still uses traditional ranking signals, and AI Overviews and other AI answer layers draw heavily on organic results. A page that doesn't rank organically is unlikely to be cited in an AI answer either — the two are connected, not separate tracks.

How do you measure success?

The same three numbers every scan reports: mention rate (how often the brand comes up in relevant queries), position within the answer, and citation share (how often a source link to your site is included). All three are published with their margin of error, per engine — see our measurement methodology.

Can small businesses compete with larger brands in GEO?

Sometimes. What we can say with evidence: in our September 15, 2026 scan our own brand was mentioned in 11.2% of ChatGPT answers (249 responses, ±3.9 points). We have no valid measurement for the other engines right now; we are waiting for a complete new scan. Scale doesn't guarantee visibility, and visibility doesn't come for free either — we're not going to claim otherwise about your brand without measuring it first.

Does Botfusions use AI to write this content?

People define the strategy and do the implementation. We use tooling for the repetitive parts — scan runs, schema generation, link and markup checks — and keep the judgment calls human.

---

Want to know where your brand actually stands before we talk about fixing anything? Get a GEO audit.