We Tested Query Fan-Out Optimization (Here's What We Learned)

Ever since Google launched AI Mode, I’ve had 2 questions connected my mind:

• How do we guarantee our contented gets shown successful AI results?
• How do we fig retired what useful erstwhile AI hunt is still mostly a mystery?

While there’s a batch of proposal online, overmuch of it is speculative astatine best. Everyone has hypotheses astir AI optimization, but fewer are moving existent experiments to spot what works.

One thought is optimizing for query fan-out. Query fan-out is simply a process wherever AI systems (particularly Google AI Mode and ChatGPT search) return your original hunt query and break it down into aggregate sub-queries, past stitchery accusation from various sources to build a broad response.

This illustration perfectly depicts the query fan-out process.

The optimization strategy is simple: Identify the sub-queries astir a peculiar taxable and past make judge your page includes contented targeting those queries. If you do that, you person amended likelihood of being selected successful AI answers (at slightest successful theory).

So, I decided to tally a mini trial to spot if this really works. I selected 4 articles from our blog, had them updated by a squad personnel to reside applicable fan-out queries, and tracked our AI visibility for 1 month.

The results? Well, they uncover immoderate absorbing insights astir AI optimization.

Here are the cardinal takeaways from our experiment:

Key Takeaways

• Optimizing for fan-out queries importantly increases AI citations: In our mini sample of 4 articles, we much than doubled citations successful tracked prompts from 2 to five. While the absolute numbers are mini fixed the sample size, citations were the main metric we aimed to influence, and the summation is directionally suggestive of success.
• AI citations tin beryllium unpredictable: I checked successful periodically during the month, and astatine 1 point, our citations went arsenic precocious arsenic 9 earlier dropping backmost down to five. There person been reports of ChatGPT drastically reducing citations for brands and publishers crossed the board. It conscionable shows really quickly things tin alteration erstwhile you're relying connected AI platforms for visibility.
• Our marque mentions dropped for tracked queries, and truthful did everyone else's: Overall, we noticed less marque references appearing successful AI responses to the queries we were monitoring. This affected our stock of voice, marque visibility, and full mention metrics. Other brands besides knowledgeable akin drops. This appears to beryllium a chopped rumor from citation changes—more astir really AI platforms handled marque mentions during our research period.

We’ll talk the results of this research successful item later successful the article. First, fto maine locomotion you done precisely really we conducted this experiment, truthful you tin understand our methodology and perchance replicate aliases amended upon our approach.

How We Ran the Query Fan-Out Experiment

Here’s really we group up and ran our experiment:

• I selected 4 articles from our blog
• For each selected article, I researched 10 to 20 fan-out queries
• I collaborated pinch Tushar Pol, a Senior Content Writer connected our team, to thief maine execute the contented changes for this experiment. He edited the contented successful our articles to reside arsenic galore fan-out queries arsenic possible.
• I group up search for the fan-out queries truthful we could measurement earlier and aft AI visibility. I utilized the Semrush Enterprise AIO level for this. We were chiefly willing successful seeing really our contented changes impacted visibility successful Google's AI Mode, but our optimizations could besides boost visibility connected different platforms for illustration ChatGPT Search arsenic a broadside effect, truthful I tracked capacity location arsenic well.

Let’s return a person look astatine each of these steps.

1. Selecting Articles

I had circumstantial criteria successful mind erstwhile selecting the articles for this experiment.

First, I wanted articles that had unchangeable capacity complete the past mates of months. Traffic has been volatile lately, and testing connected unstable pages would make it intolerable to show whether immoderate changes successful capacity were owed to our modifications aliases conscionable normal fluctuations.

Second, I avoided articles that were halfway to our business. This was an experiment, aft all. If thing went wrong, I didn't want to negatively impact our visibility for captious topics.

After reviewing our contented library, I recovered 4 cleanable candidates:

1. A guideline connected really to create a trading calendar
2. An explainer connected what subdomains are and really they work
3. A broad guideline connected Google keyword rankings
4. A elaborate walkthrough connected really to behaviour method SEO audits

2. Researching Fan-Out Queries

Next, I moved connected to researching fan-out queries for each article.

There's presently nary measurement to cognize which fan-out queries (related questions and follow-ups) Google will usage erstwhile personification interacts pinch AI Mode, since these are generated dynamically and tin alteration pinch each search.

So, I had to trust connected synthetic queries. These are AI-generated queries that approximate what Google mightiness make erstwhile group hunt successful AI Mode.

I decided to usage 2 devices to make these queries.

First, I utilized Screaming Frog. This instrumentality fto maine tally a custom script against each article. The book analyzes the page content, identifies the main keyword it targets, and past performs its ain type of query fan-out to propose related queries.

Unfortunately, the information isn’t decently visible wrong Screaming Frog—everything sewage crammed into a azygous cell. So, I had to transcript and paste the full compartment contents into a abstracted Google Sheet.

Now I could really spot the data.

The bully point is that the book besides checks whether our contented already addresses these queries. If immoderate queries were already addressed, we could skip them. But if location were caller queries, we needed to adhd caller contented for them.

Next, I utilized Qforia, a free instrumentality created by Mike King and his squad astatine iPullRank.

The logic I utilized different instrumentality is simple: Different devices often aboveground different queries. By casting a wider net, I'd person a much broad database of imaginable fan-out queries.

Plus, if definite queries are communal crossed some tools, that's a awesome that addressing them whitethorn beryllium important.

The measurement Qforia useful is straightforward: Enter the article's main keyword successful the fixed field, adhd a Gemini API key, prime the hunt mode (either Google AI Mode aliases AI Overview), and tally the analysis. The instrumentality will make related queries for you.

After moving the study for each article, I saved the results successful the aforesaid Google Sheet. 

3. Updating the Articles 

With a spreadsheet afloat of fan-out queries, it was clip to really update our articles. This is wherever Tushar stepped in.

My instructions were simple:

Check the fan-out queries for each article and reside those that weren’t already covered and were feasible to add. If immoderate queries felt for illustration they were beyond the article's scope, it was OK to skip them and move on.

I besides told Tushar that including the queries verbatim wasn't ever necessary. As agelong arsenic we were answering the mobility posed by the query, the nonstop wording didn't matter arsenic much. The extremity was making judge our contented included what readers were really looking for.

Sometimes, addressing a query meant making mini tweaks—just adding a condemnation aliases 2 to existing content. Other times, it required creating wholly caller sections.

For example, 1 of the fan-out queries for our article astir doing a method SEO audit was: "difference betwixt method SEO audit and on-page SEO audit." 

We could’ve addressed this query successful galore ways, but 1 smart action was to make a comparison correct aft we specify what a method SEO audit is.

Sometimes, it wasn't easy (or moreover possible) to merge queries people into the existing content. In those cases, we addressed them by creating a caller FAQ conception and covering aggregate fan-out queries successful that section.

Here’s an example:

Over the people of 1 week, we updated each 4 articles from our list. These articles didn't spell done our modular editorial reappraisal process. We moved fast. But that was intentional, fixed this was an research and not a regular contented update.

4. Setting Up Tracking

Before we pushed the updates live, I recorded each article’s existent capacity to found a baseline for comparison. This way, we would beryllium capable to show if the query fan-out optimization really improved our AI visibility.

I utilized our Enterprise AIO level to way the results. I created a caller task successful the instrumentality and plugged successful each the queries we were targeting. The instrumentality past began measuring our existent visibility successful Google AI Mode and ChatGPT.

Here’s what capacity looked for illustration astatine the commencement of this experiment:

• Citations: This measures really galore times our pages were cited successful AI responses. Initially, only 2 retired of our 4 articles were getting cited astatine slightest once.
• Total mentions: This metric shows the ratio of queries for which our marque was straight mentioned successful the AI response. That ratio was 18/33—meaning retired of 33 tracked queries, we were being mentioned for 18 queries.
• Share of voice: This is simply a weighted metric that considers some marque position and mention wave crossed tracked AI queries. Our people was 23.4%, which indicated we were coming successful immoderate responses but not each aliases successful the lead positions.
• Brand visibility: This told america what percent of punctual responses mentioned our marque astatine slightest once, sloppy of the position.

I decided to hold 1 period earlier logging metrics again. Then, it was clip to reason our experiment.

The Results: What We Learned About Query Fan-Out Optimization

The results were honestly a mixed bag.

First off, immoderate bully news: our full citations increased.

Our 4 articles went from being cited 2 times to 5 times—a 150% increase. For example, 1 of the edits we made to the method SEO article (which we showed earlier) sewage utilized arsenic a root successful the AI response.

Seeing our contented cited is precisely what we hoped for, truthful this is simply a win. (Despite the mini sample size.)

Interestingly, our last results could’ve been much awesome if we ended our research earlier. At 1 point, we sewage to 9 citations, but past they decreased erstwhile ChatGPT significantly reduced citations for each brands. 

This conscionable shows really unpredictable AI platforms tin be, and that factors wholly extracurricular your power could effect your visibility.

But what astir the different metrics we tracked?

Our stock of sound went down from 23.4% to 20.0%, marque visibility fell from 13.6% to 10.6%, and our marque mentions dropped from 18 to 10.

According to our data, we're not the only ones who saw declines successful marque metrics. Here's a floor plan showing really galore brands’ stock of sound went down astatine the aforesaid time.

This happened because AI platforms mentioned less marque names wide erstwhile generating responses to our tracked queries. This was a wholly different rumor from the citation fluctuations I mentioned earlier.

Considering the outer factors, I judge our optimization efforts performed amended than the information shows. We managed to summation our citations contempt the things moving against us.

So, now the mobility is:

Does Query Fan-Out Optimization Work?

Based connected what we learned successful our experiment, I'd opportunity yes—but pinch a immense asterisk. 

Query fan-out optimization tin thief you get much citations, which is valuable. But it’s difficult to thrust predictable maturation erstwhile things are this volatile. Keep this successful mind erstwhile you’re optimizing for AI.

If you’re willing successful learning much astir AI SEO, support an oculus retired for the caller contented we regularly people connected our blog. Here are immoderate articles you should cheque retired next:

• How AI Search Really Works: Findings from Our AI Visibility Study
• What Is LLMs.txt & Should You Use It?
• Why Your Brand Is Your Most Important SEO Asset successful 2026