Every Google Ads account contains some degree of wasted spend.
The obvious examples are usually easy to find. Search terms containing words such as “free”, “jobs” or other clearly irrelevant phrases can often be excluded quickly.
The more difficult waste is hidden in patterns across hundreds or even thousands of search terms.
Individual searches may only receive one or two clicks, meaning there is not enough data to confidently exclude them. However, when you analyse the words and phrases appearing repeatedly across those search terms, you can uncover patterns that would be almost impossible to identify manually.
This is where N-Gram analysis can be particularly useful.
By combining your Google Ads search term data with AI, you can aggregate performance across recurring words and phrases, identify where spend is consistently failing to generate conversions and uncover potential negative keyword opportunities at scale.
Why Traditional Search Term Analysis Has Limitations
Search term analysis is one of the most important optimisation activities in Google Ads.
The search terms report shows you what users actually typed into Google before triggering and clicking your adverts. You can compare these searches with your keywords, identify irrelevant traffic and add negative keywords to prevent similar searches from triggering your ads again.
Every pound spent on genuinely irrelevant traffic is money that could potentially have been spent on searches with a greater chance of converting.
The problem is scale.
Large Google Ads accounts can generate enormous search term reports. Some accounts can accumulate hundreds of thousands of different searches in a relatively short period.
Most of those individual search terms may only receive one or two clicks.
You cannot reasonably exclude every search term that fails to convert after two clicks because there simply is not enough data to determine whether it is genuinely poor traffic.
As a result, you can end up with thousands of low-volume search terms that individually tell you very little.
Collectively, however, they may contain extremely useful patterns.
What Is N-Gram Analysis?
N-Gram analysis involves breaking search terms into individual words and combinations of words, then analysing how those components perform across a larger dataset.
Take the phrase:
free trial
You can break this down into:
- “free”
- “trial”
- “free trial”
The individual words are one-grams, while “free trial” is a two-gram.
When you repeat this process across an entire Google Ads search term report, you can aggregate performance for every recurring word and phrase.
Instead of asking whether one individual search term converted, you can begin asking questions such as:
- How much have we spent on all search terms containing this word?
- How many clicks have searches containing this phrase generated?
- How many conversions have searches containing this N-Gram produced?
- Are there certain recurring words associated with unusually poor performance?
- Are there phrases that consistently appear within non-converting searches?
This can reveal patterns that are almost impossible to spot by manually scrolling through the Google Ads interface.
Why N-Gram Analysis Matters More in Modern Google Ads
Search matching has become significantly broader.
Features such as:
- Broad match
- Performance Max
- AI Max
- Automated query matching
can allow Google to match your advertising with searches that extend considerably beyond the precise wording of your original keywords.
This can be extremely useful.
Google may discover converting searches that you would never have thought to target manually.
However, broader matching also creates more opportunities for wasted spend.
The larger and more diverse your search term data becomes, the harder it is to identify those patterns manually. N-Gram analysis gives you a method of analysing that expanded search traffic at scale.
Using AI to Analyse Google Ads Search Terms
Historically, this type of analysis would usually require a Google Ads script, specialist software or relatively advanced spreadsheet work.
AI tools can make the process considerably easier.
The example in the video uses Claude, but the important part is the underlying process: provide the AI model with clear instructions, upload a sufficiently large search term report and ask it to aggregate the performance of recurring N-Grams.
Step 1: Export Enough Search Term Data
Start inside Google Ads and navigate to your search terms report.
Use a sufficiently long date range.
N-Gram analysis depends on aggregation, so analysing only a few days of activity may produce very little useful information. You want enough search volume for repeated words and phrases to accumulate meaningful click and cost data.
For higher-volume accounts, you may already have sufficient data across a relatively short period. Lower-volume accounts may require several months.
Export the report as a CSV file.
Include relevant metrics such as:
- Search term
- Impressions
- Clicks
- Cost
- Conversions
- Cost per conversion
- Conversion value
- ROAS, where applicable
The more relevant performance data you provide, the better the analysis can be.
Step 2: Give the AI Clear Instructions
The AI needs to understand what you are trying to achieve.
Your instructions should explain that the purpose of the exercise is to break search terms into N-Grams, aggregate their performance and identify words or phrases that may represent potential wasted spend.
It is also important to specify minimum data thresholds.
For example, you might initially tell the model not to flag an N-Gram unless it has accumulated at least:
- 10 clicks
- £100 in spend
- Or another threshold appropriate for the account
These figures should not be treated as universal rules.
The right threshold depends heavily on your business and typical cost per conversion.
If your normal cost per acquisition is £200, a phrase spending £100 without converting may not be unusual at all.
How to Review the N-Gram Analysis
Once the AI has processed the data, you can review the N-Grams by metrics such as cost and conversions.
A useful starting point is to filter for N-Grams with:
Zero conversions and the highest total spend.
This immediately draws your attention towards the biggest potential opportunities.
For example, the analysis demonstrated in the video identified phrases including “home design” and “services” that had generated meaningful numbers of clicks and measurable spend without producing conversions during the analysed period.
These were particularly interesting because they were not obviously irrelevant phrases.
They appeared commercially relevant to the business.
That is exactly why this method can be useful.
A human reviewing a normal search term report would be unlikely to notice that a common phrase was repeatedly appearing across many different searches without converting. By aggregating those searches through N-Gram analysis, the pattern becomes visible.
However, that does not mean the phrase should automatically be added as a negative keyword.
Do Not Treat AI Recommendations as Automatic Decisions
This is probably the most important part of the process.
The AI should perform the analysis. You should make the final decision.
An N-Gram having zero conversions does not necessarily make it a bad keyword.
Imagine your normal cost per conversion is £200 and a particular N-Gram has accumulated £80 in spend without converting.
That is not enough evidence to conclude that the traffic is poor.
Even the 10-click or £100 thresholds used during an initial analysis may be too low for many advertisers.
When reviewing a potential exclusion, consider:
- Normal account CPA
- Typical conversion rate
- Average order value
- Conversion lag
- Total spend attached to the N-Gram
- Number of clicks
- Commercial relevance
- How much traffic could be lost if the phrase is excluded
You should generally set a relatively high standard before excluding commercially relevant phrases.
Saving £100 of wasted spend is not valuable if the negative keyword accidentally prevents £1,000 of future revenue.
Be Careful With Overlapping Data
There is another important technical consideration.
N-Gram totals are not necessarily additive.
One search term may contribute data to several different N-Grams.
For example, a three-word search could contribute spend to each individual word, multiple two-word combinations and the complete three-word phrase.
If several N-Grams are flagged, you cannot simply add all of their reported costs together and assume that represents your total potential saving.
The same underlying search term spend may appear across multiple rows.
Use the report to identify patterns, not to calculate potential savings by simply adding every flagged N-Gram together.
Use a Separate Negative Keyword List
If your analysis identifies genuine negative keyword opportunities, it is worth implementing them carefully.
A particularly sensible approach is to create a separate negative keyword list within the Google Ads Shared Library specifically for exclusions discovered through N-Gram analysis.
Attach that list to the relevant campaigns.
There is a major advantage to doing this.
If the exclusions reduce traffic or damage campaign performance, you can quickly detach the entire list and effectively reverse the change.
If you instead mix dozens of new negatives into your existing campaign or account-level lists, reversing the optimisation becomes much more difficult.
You may need to examine change history, identify every keyword that was added and manually remove them again.
A dedicated list gives you considerably more control.
Repeat N-Gram Analysis Over Time
N-Gram analysis should not necessarily be a one-off exercise.
Search behaviour changes, Google continues expanding its matching technology and your account continually accumulates more data.
Running the analysis periodically can reveal patterns that were not statistically meaningful the previous time you checked.
As more clicks and spend accumulate, previously uncertain N-Grams may become much clearer candidates for exclusion.
The same dedicated negative keyword list can then be expanded gradually as new evidence appears.
Final Thoughts
Traditional search term analysis remains essential, but manually reviewing individual searches becomes increasingly difficult as Google Ads accounts grow.
N-Gram analysis provides another layer of insight.
By breaking search terms into recurring words and phrases and aggregating their performance, you can identify patterns of wasted spend that are hidden across hundreds or thousands of individual searches.
AI makes that analysis considerably more accessible.
The key, however, is not to blindly accept its recommendations.
Export enough data, use thresholds appropriate to your account, focus on the N-Grams with the most meaningful levels of spend and manually review every potential negative keyword before making changes.
If you do decide to implement exclusions, keep them in a separate negative keyword list so the optimisation can be reversed easily if performance deteriorates.
Used carefully, N-Gram analysis can help uncover wasted Google Ads spend that conventional search term analysis may never reveal.
