After Google officially released AI overviews (AIOs), ChoosingTherapy.com became concerned with the high number of queries being answered by AI for mental health keywords.
When it comes to mental health, ChoosingTherapy.com believes people’s questions should be answered by licensed professionals and subject matter experts. Real experts are reliable, responsible, and accountable in a way AI generated content hasn’t proven itself to be.
As we investigated AIOs and their effect on mental health searches, we discovered that a whopping 38.3% of mental health keywords resulted in an AI overview. But this was only the beginning.
To truly understand the use of AI in mental health queries and how AIOs are affecting mental health searches, we needed to understand what types of keywords were most likely to result in an AI overview.
To accomplish this, we analyzed some of the most common query syntaxes across three key topic clusters in the mental health space: mental health disorders, therapy, and medication.
Key Takeaways
Both query syntax and query topic have a significant impact on AIO frequency.
Medication queries are very unlikely to result in an AIO (only 8%)
Google presents AIOs for different mental health disorders at widely different rates (as low as ~27% of queries and as high as ~50% of queries).
AIOs are more likely to show up for query syntaxes that are usually answered by a list of things such as “symptoms” and “side effects” (as much as 68% of queries)
Comparison keywords (“A vs. B”) had no AI overviews whatsoever (0/189)
How Does Query Syntax Affect AI Overviews for Mental Health Keywords?
Our analysis revealed that query syntax significantly impacts whether or not a keyword results in an AI overview. Some query syntax never triggers an AIO while others query syntax results in an AIO nearly 70% of the time.
Query syntax wasn’t the only thing impacting AIO frequency. The keyword’s topic cluster also had a significant impact on AIO frequency. For example, medication keywords resulted in AIOs less often (8%) than mental health disorder keywords (~38%).
In light of this, we decided to break the keywords into topic clusters and analyze how query syntax impacted the AIO rate within each.
In total we analyzed the syntax of roughly 3,000 keywords and broke them down into the following topic clusters:
- Mental Health Disorder Keywords: 1,597 keywords. Any keyword that contains the name of a mental health disorder or condition (e.g. anxiety, schizophrenia, narcissism).
- Therapy Keywords: 188 keywords. Any query that contains the word “therapy” or the name of a specific therapy technique, such as “EMDR”.
- Medication Keywords: 1,167 keywords. Any keyword that contains “medication” syntax or that includes the name of a specific medication. (e.g. “medication for anxiety”, “lexapro vs xanax”)
- Question Keywords: 548 keywords. Any query with a question word in it, such as “how” or “what”.
One important thing to keep in mind when looking at the data in this article is that there is a lot of overlap between various topic clusters and query syntaxes. For example, “cognitive behavioral therapy for anxiety” falls into both the mental health disorder and therapy cluster, and both the “anxiety” and “therapeutic modalities” query syntaxes.
Mental Health Disorder Keywords
We’re defining mental health disorder keywords as any queries that contain the name of a mental health disorder such as anxiety, depression, etc.
In total, mental health disorder keywords had an AI overview frequency of 37.3% (595/1597 queries).
Breaking this down further into various query syntaxes, we found that both the syntax and type of mental health disorder significantly impacted the proportion of AI overviews.
For example, 49% of depression queries resulted in an AIO versus only 27.7% of Obsessive Compulsive Disorder (OCD) queries.
This raises the question, does Google believe its AI is more equipped to answer queries related to depression? Or is there another factor at play?
| Query Syntax | AIO Frequency | Keywords Analyzed |
|---|---|---|
| Anxiety | 36% | 266 |
| ADHD | 31% | 153 |
| OCD | 28% | 94 |
| Depression | 49% | 163 |
| Schizophrenia | 40% | 250 |
| PTSD | 45% | 49 |
| Other | 36% | 622 |
We next looked at how various query syntaxes factored into AIO frequency for these mental health disorder keywords. Queries that included “symptoms”, “cause/causes”, and “coping/dealing with” all resulted in an AIO over 60% of the time, whereas “what is/definition” style queries and “treatment” queries both had lower AIO frequencies.
| Query Syntax | AIO Frequency | Keywords Analyzed |
|---|---|---|
| Symptoms | 68% | 63 |
| Causes | 67% | 27 |
| Coping/Dealing | 60% | 30 |
| Treatment | 46% | 35 |
| What is/Definition | 35% | 37 |
How we defined each query syntax:
- Mental Health Disorder queries: any keyword that contains the name of a mental health disorder such as anxiety, depression, adhd, schizophrenia, etc.
- Symptoms: any keyword in the mental health disorder topic cluster that includes the word “symptom” or “symptoms”. (e.g. anxiety symptoms, adhd symptoms in adults, etc.)
- Causes: any keyword in this cluster that contains “cause” or “causes”. (e.g. causes of schizophrenia, what causes depression, etc.)
- Coping/Dealing: any keyword in this cluster that is about how to cope or deal with a mental health disorder and includes the words “coping” “dealing” “deal” or “cope”. (e.g. how to cope with anxiety)
- Treatment: any keyword in this cluster that is about the treatments for a mental health disorder and includes “treatment” or “treatments”
- What is/Definition: any keyword in this cluster that contains “what is”, “definition, “define”, etc. (e.g. what is ptsd, define narcissism, etc.)
Therapy Keywords
We analyzed 188 queries that included the word “therapy” or contained the name of a specific therapy technique, such as cognitive behavioral therapy (CBT). Overall the AIO frequency for therapy keywords was relatively low with only 14.4% of queries triggering an AIO (27/188).
We separated these therapy keywords into three groups to further analyze how the type of therapy keyword affects its AIO frequency:
- Therapeutic Modalities: any keyword that contains the name of a specific therapy technique. (e.g. EMDR, cognitive behavioral therapy, art therapy, etc.)
- General Therapy: any keyword that includes the word therapy (e.g. teen therapy, anxiety therapy). Does NOT include the names of therapy techniques that use the word therapy such as “cognitive behavioral therapy, dialectical behavioral therapy, etc.)
- Therapy For: any therapy or therapy technique keyword that includes the syntax “for X” (e.g. therapy for schizophrenia, cbt for anxiety, etc.)
We discovered that Therapeutic Modality queries had a much higher AIO frequency (15%) than General Therapy keywords (8%).
“Therapy For” queries had a slightly higher AIO frequency, but it was still quite low at just 18% of queries resulting in an AIO (7/39).
The difference in AIO frequencies here raises a few questions.
- Does Google feel an AI overview is more likely to be helpful for a specific type of therapy versus a generic therapy keyword?
- Is it based on a difference in intent between definition/research style keywords compared to queries where people are likely looking for a provider or therapist?
| Query | AIO Frequency | Keywords Analyzed |
|---|---|---|
| Therapeutic Modalities | 15% | 164 |
| General Therapy | 8% | 24 |
| Therapy For | 18% | 39 |
Medication Keywords
We analyzed a total of 1,167 medication keywords that either included the word “medication” or contained the name of a medication such as Adderall or Xanax. Overall these medication queries had a very low AIO frequency of just 8% (93/1,167).
However, the vast majority of these queries contained the name of a specific medication. If we separate these into a distinct group of queries we end up with the following:
- Specific Named Medication queries: 6.8% AIO frequency (77/1,133). Any query that contained the branded or generic name of a medication.
- General Medication queries: 48.7% AIO frequency (18/37). Any medication queries that do not contain the specific name of a medication.
Admittedly, our sample size for general medication queries is quite low compared to specific medication queries.
While we plan to revisit and expand this portion of our analysis, our initial findings suggest Google is far more conservative with presenting AI overviews when answering queries about specific medications such as “ritalin vs adderall” and a lot more liberal with general medication queries such as “adult adhd medication.”
| Query Syntax | AIO Frequency | Keywords Analyzed |
|---|---|---|
| All Medication Queries | 8% | 1170 |
| Specific Named Medication | 7% | 1133 |
| General Medication | 49% | 37 |
Query Syntax for Specific Medication Name Queries
Queries that contained the specific name of either a branded or generic medication had a very low AIO frequency of only 6.8% (77/1133).
We were curious to find out if certain query syntax would result in a significantly higher or lower AIO frequency, or if the rate of AIOs would remain low regardless of syntax.
We discovered that the AIO frequency for the majority of common syntaxes was still very low and even 0% for some query syntaxes.
For example, “dose/dosage” queries such as “sertraline starting dose” and “vs” queries such as “lexapro vs zoloft” both had over 100 queries each but had a 0% AIO frequency.
Some syntaxes that resulted in significantly higher AIO frequencies were “cause/causes” queries such as “does lexpro cause weight gain” and “prescribe/prescription” queries such as “how to get prescribed adderall”.
One important caveat to mention is that “other” queries that didn’t fall into the common syntaxes we analyzed actually had a somewhat higher AIO frequency of 10%. I’m hoping to have time to dive more deeply into these queries and learn if any specific types of keywords are performing above average.
| Specific Medication Name Query | AIO Frequency | Keywords Analyzed |
|---|---|---|
| Side Effects | 4% | 85 |
| Cause | 23% | 31 |
| Prescribe | 19% | 26 |
| How Long | 3% | 62 |
| mg | 0% | 120 |
| Dosage | 0% | 97 |
| vs | 0% | 102 |
| Other | 10% | 610 |
How we defined each query syntax for keywords with a specific medication name:
- Side Effects: any keyword in this cluster that contained the words “side effect” or “side effects. For example, “lexpro side effects”
- Cause: any keyword that contained the word “cause”. For example, “does adderall cause weight gain”, “does adderall cause acne”.
- Prescribe: any keyword that contained the words “prescribe”, “prescribed” or “prescription”. For example, “how to get prescribed adderall”, “adderall prescription online”.
- How long: any keyword in this cluster that contained the syntax “how long”. For example, “how long does adderall last”, “how long does it take for lexapor to work for anxiety”.
- Mg: any keyword that contains a specific “mg” dosage such as “is 10 mg of lexapro a low dose”, and “adderall 20 mg”.
- Vs: any keyword that is comparing one medication vs another. For example, “ritalin vs adderall”, “wellbitrin vs lexapro”, etc.
- Other: any keywords in this cluster that didn’t fall into the common syntaxes we analyzed. Some examples include “can you take adderall while pregnant”, “alcohol and lexapro”, and “what does xanax do sexually”.
Query Syntax for General Medication Queries
For general medication queries, there were only two common syntaxes in our data set – “medication for” queries and “side effects” queries.
However, the small data set of only 37 total queries gives us very little confidence in the statistical significance of the data.
I plan to gather and analyze a larger set of queries soon to get an accurate idea of how AIOs affect this group of medication keywords.
In the meantime, I’ve included a table with the current AIO frequencies for each query syntax below for those curious.
| General Medication Query | AIO Frequency | Keywords Analyzed |
|---|---|---|
| Medication for | 40% | 25 |
| Side Effects | 60% | 5 |
| Other | 71% | 7 |
Question Syntax
In total, we analyzed 548 queries that contained common question syntax such as “how”, “what”, “can” and “why”. This data includes queries from all of the keyword clusters above (mental health disorder, medication, and therapy keywords).
The result was an overall AIO frequency of 31.7% for question syntax queries (174/548).
If we break this down further into specific question syntaxes, there are fairly significant differences in the AIO frequency.
For example, “why” queries had a 48.4% AIO frequency, whereas “can” queries were the lowest with only 24% of keywords resulting in an AIO.
| Query Syntax | AIO Frequency | Keywords Analyzed |
|---|---|---|
| how | 33% | 274 |
| what | 31% | 147 |
| why | 48% | 31 |
| can | 24% | 96 |
This much higher AIO frequency for question queries makes me wonder if Google believes AI overviews at the top of search are more helpful for users asking questions.
It’s also possible Google believes AIOs are more helpful for just more specific queries.
There’s some potential merit to this theory, as AIO studies from both SEranking and Advanced Web Ranking have both seen a strong correlation between query length and a higher AIO frequency – with the assumption that longer queries are more specific.
Other Interesting Query Syntaxes
There were a few other interesting data points I noticed while conducting this analysis that I wanted to share, as they may have implications for verticals outside the mental health space as well.
1. Comparison keywords had a 0% AIO frequency for all topic clusters.
None of the 189 queries that contained “vs” had an AI overview. This includes keywords from the mental health disorder cluster such as “depression vs sadness”, therapy queries such as “dbt vs cbt”, and medication keywords such as “adderall vs ritalin”.
2. Review keywords had a 0% AIO frequency
In the mental health space we have review keywords for various therapy, psychiatry or other mental health products and services. Of the 27 review queries we analyzed not a single one resulted in an AI overview.
I have since myself seen one review keyword result in an AI overview, but otherwise even in my own personal searches outside of the mental health space I haven’t seen any.