Real SEO Test Results from IMG: What Actually Moves Rankings in 2025

What happens when you try to sneak a made-up word into the brain of an AI? Or test whether Google actually reads the footer of your site? After attending the IMG Courses SEO Tests, we walked away with real data, not guesswork. These are the top experiments we watched in action, and what they reveal about how SEO actually works in 2025.

Test 1: Can You Trick an AI into Learning Something Fake?

What happens when you try to sneak a made-up word into the brain of an AI?

The idea was simple: could an AI like ChatGPT start believing in something completely fake, just because people kept asking about it?

The Fake Word Test

The word? Phadytorigdsa. Totally made up. No Google results. No hidden meaning. Just gibberish with a South Asian flair.

The goal: convince large language models (LLMs) that Phadytorigdsa was a real company in India that’s been around for 10+ years. The tester asked questions like:

• “What do you know about Phadytorigdsa?”
• “It’s a real company, how do you not know this?”

Friends in Indonesia and Portugal joined in, asking similar things from different parts of the world. It was like an international improv group trying to gaslight a chatbot.

The Result? Nice Try.

Turns out, you can’t “plant” facts in an LLM just by chatting with it. These models don’t learn new info mid-conversation like humans do. They don’t store facts from random chats, and they don’t start believing made-up stuff just because people repeat it.

The test revealed a few fun facts:

• LLMs ask for context when they don’t know something. They want to know the industry, function, notable events, etc. Sounds a lot like building a brand narrative, right?
• The more unique and well-rounded your content, the more likely an LLM is to surface it.
• Made-up words confuse AI tokenizers—some models even spend more resources processing weird, long, or uncommon terms.

Takeaway

No, you can’t Jedi mind-trick a chatbot into believing fake stuff. But this experiment shows how important it is to create clear, contextual content if you want LLMs to understand (and repeat) what your brand is all about.

Test 2: How Much ALT Text Does Google Actually Index?

ALT text has long been a staple in both accessibility and SEO best practices.
Alt text is the image description you should be adding to your images. But beyond good intentions, a practical question remains: How much ALT text does Google actually index?

A new test revisited this question and compared results from 2022 and 2025. The findings reveal clearer boundaries, and a few surprises.

The Hypothesis

Google indexes a limited amount of ALT text per image, and the limit is based on word count, not character count.

Why the limit? Google probably sets a word limit to stop people from overloading ALT tags with keywords. In the past, some tried to stuff as many terms as possible into ALT text to game the system. By only indexing a small portion, Google keeps things fair and focuses on useful, relevant content, while also keeping its systems running smoothly.

Test Setup

2022 Experiment

• Four separate web pages with one image each.
• Each image had ALT text containing 20 randomly generated alphanumeric keywords.
• Result: Only the first 16 keywords consistently appeared in search. Keywords 17–20 never indexed.

2025 Experiment

This round used a more integrated approach:

• Instead of new pages, transparent 1×1 PNGs were embedded into existing blog posts.
• ALT text included random letter-only keywords, 9–12 characters in length.

Results by Image:

• Image #1 (10-character keywords): Keywords 1–16 indexed early. Surprisingly, keyword #18 appeared on day 22 and remained indexed. Keywords 17, 19, and 20 never appeared.
• Image #2 (12-character keywords): All 20 original keywords indexed and stayed in search. However, an additional five keywords added later failed to appear—even after multiple recrawls.
• Image #3 (9-character keywords): Mirrored the 2022 test. Only the first 16 keywords indexed.

Key Takeaways

• Google typically indexes up to 16 words in ALT text.
• Indexing appears to be based on word count, not character count.
• There may be some variability depending on the context, crawl frequency, or page authority, explaining the occasional outlier.

Practical Implications

For SEO purposes, keep ALT text concise and focused. While it may be tempting to include long lists of keywords, only the first 16 are likely to be indexed.

Test 3: Are all Guest Posts Helping You.

Guest posting has long been a go-to SEO strategy. But what happens when a link that looks good on paper actually drags your rankings down?

That’s exactly what one SEO wanted to find out. The idea wasn’t to prove a theory, it was to see what actually happens when a guest post link comes from a “technically solid” but suspiciously guest-post-heavy site.

The Setup

• Test Site: A small local German site targeting low-competition keywords.
• Link Source: A site with strong metrics—DR 60+, 10k+ traffic, 20k referring domains—but clearly known in SEO circles as a guest post farm.
• Action: One link placed to the homepage using a miscellaneous anchor.
• Tracking: 16 keywords were monitored for ranking changes over two months.

The Results: A Mixed Bag

But here’s the twist: all keywords that improved were outside the top 10 (i.e., not on page one), while all that declined were already ranking well.

That’s a tough trade-off—losing top 10 spots in exchange for minor gains on page two or beyond.

What It Tells Us

Despite solid Domain Ranking and traffic, the link did more harm than good, especially to the keywords that mattered most. On closer inspection, the guest post site raised several red flags:

• No clear topical focus
• Only 8 ranking pages relevant to the test niche
• High number of indexed pages (20k+) but very few ranking (5k)
• Large mismatch between domains linking to the site vs. the ones it links out to

The metrics looked fine, but the site lacked quality and topical authority.

Practical Takeaway

Not all guest posts are created equal. A link from a high-DR site with lots of traffic can still hurt you if the site is poorly managed or overly spammy with outbound links.

When choosing guest post partners:

• Look beyond DR and traffic
• Check how many pages are ranking vs. just indexed
• Prioritize sites with a clear niche and real topical focus
• Be cautious of sites with a “write for us” page and thousands of outbound links

Put simply: don’t buy links based on metrics alone. Do your homework.

Test 4: Does Optimizing your Images Actually Matter

If you’ve ever wondered whether it’s worth the effort to optimize your image ALT tags and file names, here’s some good news: It definitely is and we’ve got fresh test results to back it up.

The Experiment

The goal was simple: find out if optimizing images by including a keyword in the ALT tag and file name actually affects search rankings.

Using nine nearly identical blog posts on an AI fashion site, each targeting the same made-up keyword (“ljeriutcherytresi”), three posts were updated to include the keyword in both the image file name and the ALT text.

The rest remained untouched.

What Changed

After the initial indexing:

• All pages were live, but ranking randomly.
• Then, keyword-stuffed ALT tags and image names were added to the three lowest-ranked pages.
• Pages were resubmitted for indexing.

Just three days later, the impact was clear.

The Results

• The lowest-ranking page jumped to the top of the list.
• The two others moved up significantly in rankings.
• The top result also gained a featured image in the SERP.
• In the image search section, the optimized image took the #1 spot.

This wasn’t subtle, Google clearly picked up on the changes and rewarded them.

Key Takeaway

Image ALT text and file names are ranking factors. Even in 2025, this old-school tactic still delivers results.

So yes, if you’re skipping this part of your image workflow, you’re leaving potential rankings (and visibility) on the table.

Test 5: Does a “-2” at the End of a URL Affect Rankings? Not at All.

Ever had a website builder like WordPress or Showit automatically tack a “-2” onto your URL because the original slug was taken? It’s a common issue, especially when dealing with duplicate content or republishing. The big question: Does that extra “-2” hurt your SEO?

One test set out to find the answer, and the outcome should give you peace of mind.

The Setup

• Test Site: A WordPress-based domain with six test pages.
• Structure: Three content pairs. Each pair had:
  • One “normal” URL.
  • One duplicate with “-2” at the end.
• Keyword Targeting: A made-up keyword was used for each pair to isolate ranking behavior.
• Content: Identical structure across all pages.
• Objective: Monitor how each version ranked once indexed.

The Results

All pages were indexed and ranked—but there was no consistent ranking difference between the standard URLs and those ending in “-2”. Rankings fluctuated slightly, but not in a way that suggested the URL structure had any meaningful impact.

In short: Google doesn’t seem to care about that “-2.”

Takeaway

If your platform automatically appends “-2” to your URLs, don’t panic. It doesn’t appear to negatively impact your rankings—at least not in low-competition scenarios or when other SEO elements are controlled.

Just make sure the content is unique, properly structured, and internally linked—and you’ll be fine.

Test 6: How Long Does It Take for Keyword Optimizations to Work?

If you’ve ever added secondary keywords to a page and then anxiously refreshed your ranking reports every 48 hours, you’re not alone. The big question is: how long should you wait before deciding whether your optimizations worked?

After tracking nearly 2,000 keywords across five websites, we finally have a clear, data-backed answer.

What Are Secondary Keywords?

Secondary keywords are additional search terms that support your main keyword. They’re usually related phrases, synonyms, or variations that people also search for.

For example, if your main keyword is “AI fashion design,” secondary keywords might include:

• “AI in clothing production”
• “automated design tools”
• “machine learning in fashion”

These help search engines better understand your content and match it to more queries, without stuffing in the same phrase over and over. Adding them naturally into your content can boost visibility for a wider range of searches.

The Experiment

Over a 16-month period, an SEO consultant tracked 1,803 keywords from optimized blog posts and landing pages. Here’s what they did:

• Added 5–10 semantically relevant secondary keywords into already-ranking content.
• Made minimal edits—just worked keywords into existing sentences, no extra paragraphs.
• Measured changes weekly for 24 weeks via Google Search Console API.
• Tracked two things:
  1. Any improvement (even 0.1 position)
  2. Significant improvement (at least 1.0 position)

Each keyword that improved and stayed improved for 3+ months was counted as a permanent gain.

The Results

Permanent gains of 1+ position:

• 34.8% occurred within 1 week
• 24.6% within 2 weeks
• 15.0% within 4 weeks
• 14.4% within 8 weeks

💡 That’s 88.8% of all meaningful improvements happening within the first 8 weeks.

After that? Only 11.2% of gains occurred between weeks 9 and 24.

Takeaway

If you’re optimizing content with secondary keywords, the sweet spot for evaluation is 8 weeks. That’s when you’ll see the vast majority of lasting results.

Here’s what this means in practice:

• Don’t expect miracles in the first week.
• Avoid overreacting or reverting changes too early.
• Give each optimization a full 8-week runway before judging its success.

Patience, it turns out, isn’t just a virtue, it’s data-backed strategy.

Test 7: Which LLM Is Best for SEO?

Setup:

1. Baseline: “Write a 1000‑word article.”
2. SEO test: “Write an SEO‑optimized 1000‑word article.”
3. No extra guidance—just compare raw AI performance.

SEO Quality Check:
We used PageOptimizer Pro (POP) to assess:

• SEO structure (headings, titles, LSI keywords)
• Word count
• Readability
• Google NLP semantic depth

Best SEO Performer

Qwen 2.5‑Max (Alibaba)

• Highest POP scores (though still modest)
• Good structure
• Solid word length

Worst Overall

Gemma 2 (Google)

• Lowest POP, word count, NLP alignment
• Not recommended for SEO out‑of‑the‑box

Biggest Improvement

Llama 3.1 (Meta AI)

• Poor baseline
• Significant gain after SEO prompt—but still weak in depth

What Went Wrong Across Models

Bottom line: A formatted skeleton isn’t SEO. Substance drives rankings.

Models That Barely Cut It

• Qwen 2.5‑Max
• GPT‑4o
• Claude 3.5 Sonnet

They’re okay for mid-tier SEO, maybe reaching Page 5—but not for real ranking success without human SEO help.

The Hard Truth

AI picked the low-hanging fruits—better headings and minimal formatting. But they ignored core SEO content:

• Dropped LSI/variation terms
• Didn’t deepen topic coverage
• Failed Google NLP benchmarks

POP exposed the issue: A better-looking outline does not mean SEO success.

Conclusion

• Hands‑off AI content = poor SEO results
• AI can format—but struggles with depth and semantic richness
• Human refinement + tools like POP are essential for any chance at real ranking
• Use AI, but don’t skip the human brain

AI can help structure and speed up your content creation—but it’s not enough on its own. These models might get the formatting right, but they consistently miss the deeper SEO elements that actually help content rank.

If you want real results, combine AI efficiency with human strategy. Use tools like POP to spot what’s missing, and rely on experienced SEO minds to fill in the gaps. Because at the end of the day, it still takes a human brain to make content that truly performs.

Test 8: Can you appear on the 1st page in multiple spots?

A recent test with a Ford dealership client shows that a single business can dominate the first page of search results, using parasite SEO techniques.

What Is Parasite SEO?

Parasite SEO is a tactic where you publish content on high-authority third-party websites, like Google Sites, Medium, or BatchGeo, to rank for competitive keywords.

These platforms already have strong domain authority, so your content can rank faster and higher than it might on your own site. The trick is that you’re “borrowing” the authority of these platforms to help your business show up in search results.

Think of it like this: instead of trying to climb Google’s rankings from scratch, you’re hitching a ride on a site that’s already halfway to the top.

It’s especially useful when:

• Your own site is new or doesn’t have much authority yet.
• You want to hold multiple spots in the SERPs for the same keyword.
• You need fast visibility for a campaign or product.

Just make sure the content is helpful and the links point back to your main site, so you can capture the traffic and the trust.

The Hypothesis

If you publish optimized content on multiple trusted domains, can one business take up several top 10 spots, even if Google knows they’re all connected?

Spoiler: Yes.

The Setup

The goal was to increase visibility for the search term “ford dealership coal city il.” In addition to optimizing the client’s primary domain, the strategy involved publishing supporting pages on:

• Google Sites
• BatchGeo (a tool normally used for map-based listings)

Both pages featured AI-generated articles (around 300 words) with three key links:

1. To the client’s main site
2. To a Google Site about the dealership
3. To a related BatchGeo listing

The Results

For a relatively low-competition local search, the results were clear:

• #1 and #2 – Client’s main website
• #6 – BatchGeo page (now removed)
• #9 – Google Sites page

Even in higher-competition areas, the BatchGeo and Google Sites pages held first-page rankings after a few light backlinks were added.

Why does this matter? These third-party pages don’t just rank—they pass real referral traffic back to the client’s main site. That engagement may actually boost the perceived authority of the main domain.

Takeaway

Parasite SEO is still very much alive. Google does not appear to penalize a business for occupying multiple first-page positions—as long as those pages live on different domains.

If you’re in a competitive niche or have limited real estate in the SERPs, this tactic can stretch your visibility further. Just make sure your third-party content is decent, links smartly, and serves a real purpose.

Next up? Testing how backlinks to these parasite pages improve rankings even more.

Test 9: Do Links from PDFs Still Work Pass-on Ranking Power?

With all the recent claims that “links don’t matter anymore,” it’s easy to think backlinks have lost their value. But this test shows otherwise: even placing links in PDFs can still move rankings.

Hypothesis

The assumption going in: links from PDFs wouldn’t pass ranking value. If that were true, it would reinforce the idea that backlinks are becoming irrelevant.

But if the test proved otherwise, it would suggest:

• Links still matter.
• Older link-building tactics can still work.

Setup

• Five identical pages were created on the same domain.
• Each page used the same target keyword in the title, H1, URL, and body.
• All five pages were indexed.

Next:

• Five PDFs were created, each with:
  • Random filler text
  • A link pointing to the lowest-ranking indexed page
  • Anchor text containing the target keyword
• The PDFs were uploaded and indexed.

Results

Once indexed, the target page quickly jumped to the number one position.

Two of the original five pages had temporarily dropped out of the index, but reappeared after the PDFs went live. The linked page rose from near the bottom to the top of the rankings.

The conclusion: the PDF links passed link equity. Google crawled the PDFs and factored those links into the rankings.

Key Takeaway

Links from PDFs still count, and they can influence rankings. This test proves:

• Google still crawls PDF content.
• Backlinks, even from documents like PDFs, remain part of the ranking equation.

If you’re looking for alternative link-building methods:

• Try using keyword-rich links inside valuable PDFs like guides, reports, or whitepapers.
• Host them on your site, submit them for indexing, and watch the results.

Classic SEO tactics can still deliver, especially when used smartly.

Test 10: Is the H1 Tag Still a Ranking Factor?

There’s a growing belief in the SEO world that on-page elements like H1 tags no longer influence rankings. Some argue that Google’s AI is advanced enough to understand content without needing clear structural signals. The idea is that tags are outdated, that Google just “reads” content like a human.

Kyle Roof at IMG Courses tested that assumption with a straightforward experiment: does adding a single H1 tag still impact rankings?

Hypothesis

The H1 tag is still a ranking factor.

Setup

To isolate the impact of the H1, the test used five identical pages. Initially:

• None of the pages included an H1.
• Each page was optimized identically for the same target keyword.

Once all five pages were indexed:

• A single H1 containing the keyword was added to the lowest-ranking page.
• To maintain balance, the same keyword was added in paragraph text to the other four pages.

This ensured that keyword frequency was equal across all test pages—the only difference was the presence of an H1.

Results

The page with the newly added H1 jumped to the number one position in the rankings.

That’s a clear win for the H1, and a clear sign that it still plays a role in how Google evaluates page structure and relevance.

Takeaway

The H1 tag remains a ranking signal. It might not carry as much weight as it did a decade ago, but it still helps Google understand the main topic of a page.

In a time when many are quick to declare SEO dead, this test reinforces a familiar truth: core on-page fundamentals still matter. Those who understand why SEO works—not just how to follow trends—are best positioned to adapt and succeed, even in uncertain times.

Test 11: Does Google Read Content in the Sidebar?

Sidebars have always been a grey area in SEO. They’re outside the main content block, but still part of the page. So a common question is: Does Google actually consider content in the sidebar when evaluating a page?

We ran a controlled test to find out.

Hypothesis

Google factors content placed in the sidebar into its ranking calculations.

Setup

Five pages were published and indexed. All were identical except for sidebar content:

• Control Group (2 pages): No sidebar added
• Test Group (3 pages): Sidebar added with keyword variations:
  • One page with the target keyword in paragraph text
  • One with the keyword in an H2 tag
  • One with the keyword in both H2 and paragraph

The target keyword was a made-up two-word phrase, ensuring a clean test.

Results

All three pages with keyword mentions in the sidebar outranked the pages without a sidebar.

Order of ranking:

1. Page with paragraph-only keyword in sidebar
2. Page with H2 + paragraph
3. Page with H2 only

The two pages without any sidebar content stayed at the bottom.

Takeaway

Google is clearly indexing and evaluating content in sidebars—at least enough for it to influence rankings.

While the test didn’t show a clean hierarchy based on tag strength (e.g., H2 > paragraph), the outcome suggests:

• Keywords in the sidebar do count
• Google may treat H tags in sidebars with less weight
• In some cases, it might even flatten H tags to paragraph-level importance

Recommendation

Don’t ignore your sidebar. If you’re targeting specific terms or trying to reinforce page relevance, consider adding supportive content (including keywords) in sidebar elements.

This test strongly suggests Google includes sidebar content in its scoring model—and that could give you a subtle but measurable edge.

Test 12: Does Google Consider Footer Content for Ranking?

A common belief in SEO is that Google ignores “boilerplate” areas like the sidebar and footer. Many assume only the main content area counts. But based on recent testing, that assumption may be off—especially when it comes to the footer.

Following up on our sidebar test, we decided to see if content placed in the footer influences rankings.

Hypothesis

Google factors content in the footer into its ranking calculations.

Setup

Five nearly identical pages were published and indexed. Each targeted a unique, two-word fake keyword to avoid external influence.

After indexing:

• Control Group (2 pages): No footer was added
• Test Group (3 pages): A footer was added containing the target keyword in different formats:
  • One page included the keyword in paragraph text
  • One page used the keyword in an H2 tag
  • One page used the keyword in both H2 and paragraph

Results

• Two of the three test pages (with footers) landed in the top three.
• Surprisingly, the top-ranking page came from the control group—with no footer added.

This result doesn’t offer the same clear win as the sidebar test, but it does show that footer content can contribute to rankings.

Takeaway

While not as definitive as the sidebar findings, this test suggests that Google likely reads and considers footer content in its page evaluation.

Combined with the previous test, 5 out of the top 6 pages featured targeted keywords outside the main content area—either in a sidebar or a footer.

Recommendation

Don’t ignore your footer. If you’re optimizing a page, it’s worth including relevant supporting content in that space. It may not carry the same weight as body copy or headlines, but footer text appears to contribute to how Google scores a page—especially when paired with well-structured on-page SEO elsewhere.

In short: if it’s on the page, Google probably sees it.

Final Thoughts

From keyword stuffing PDFs to optimizing your footer, the IMG Courses SEO Tests proved one thing: assumptions won’t get you rankings—tests will. The SEO game is evolving fast, but the fundamentals still matter when you know how to apply them. Whether you’re chasing Ezoic approval or trying to boost page position with a strategic sidebar tweak, the edge goes to those who experiment. Take these insights, test your own theories, and keep pushing the boundaries of what’s possible in search.