In ancient Greece, priests gathered at the Temple of Apollo on Mount Parnassus, the better to interpret the deity-inspired prophecies of Pythia, the Oracle of Delphi.

Google, for all its omnipresence, is no deity, yet legions practice search-engine optimization in an effort to divine the secret algorithms that move certain websites to the top of an online search.

Now comes candidate optimization, essentially the same idea applied to job seekers, with services that help them tweak their resumés so they end up at the top of a list generated by a company's computerized talent-management system.

Hurdle No. 1? Dazzling the computers, which these days are more likely than people to be frontline resumé readers.

"Computers have a very specific way that they look at resumés," said Jon Ciampi, founder of California-based Preptel Corp. For a subscription of $25 a month, Ciampi's two-year-old company will help a candidate reconfigure a resumé so a computer can better spot the qualifications that make that candidate a good match for the job.

With the recession and its lingering aftermath, the number of resumés generated for any job opening is so overwhelming that human-resource staffs can't handle them all. The U.S. Labor Department reported that in February there were 3.5 million job openings and 12.8 million people unemployed.

No wonder recruiters turn to software created by companies such as Kenexa Inc., in Wayne, a leader in the field.

"The volume is just too high," said Jeffrey Weidner, director of sourcing science at Kenexa, whose website describes a case study in which one wireless-communications company received five million resumés, leading to 50 hires.

In 2011, companies worldwide spent $3 billion on talent-management systems that scan and sort resumés, schedule interviews, track follow-ups, and incorporate social media, according to Bersin & Associates L.L.C., a California consulting group that tracks the industry.

All this talent management may be efficient, but it can be maddening for job seekers, who spend endless hours applying online, only to feel as if they've sent their resumés - and their hopes - into a realm impossible to divine.

"It's very frustrating," said Keith Conrad, a laid-off administrator who lost his Philadelphia city job in July. "You know you have a good resumé. You know you have good experience, you see a good job, but you never get a chance to sell yourself." (Read about Keith Conrad's story here.)

Most counselors advise that the best way to get a job is networking, making connections through people you know. But there will remain some portion of the hunt that relies on online applications.

Given the new reality, applicants need to learn how to communicate with nonhuman recruiters. Keyword matching forms the basis of that communication, just as it does in web searches.

If the employer wants a programmer who knows Java, a computer language, it's easy enough to search resumés for the word. But included may be a resumé from a not-so-computer-literate barista foaming cappuccino at Philadelphia Java Co. at Fourth and South Streets.

When it was first developed, this type of software "could find the content in your resumé, but it didn't find the context," said Weidner, who began his career in recruiting in 1995, those days when resumés were faxed.

Over time, algorithms have become more able to find context. A computer can look for other resumé words indicating a programmer, such as other computer languages. The more matches, the higher a resumé rises on a "relevance" list, Weidner said.

And it's not only the number of matches, but how far apart certain words are. For a programmer, it might be better to have Java near the word protocol, not cashier.

Ciampi said job seekers really need to prepare two categories of resumés - one for humans, the other for computers. For example, he said, a resumé for humans might use the words career successes and accomplishments, but a computer scanner would rather see work experience.

Work history presented in a spreadsheet is unreadable for a computer, no matter how clear it is to a human. Computers, Weidner said, like to see company name, followed by title, dates, and description of duties. Varying the format may confuse a computer.

Ciampi, once an executive at a talent-management software company, said Preptel's Resumeter service can let applicants know how closely their resumés match a job posting, so they can know how hard to pursue it.

Only the top 10 percent to 20 percent of resumés, those with the highest number of matches, make it through to a human for further consideration, Weidner said.

And all this assumes that the employers writing the job descriptions are actually capable of crafting postings that clearly reflect the skills and qualities needed for the jobs.

Conrad, the laid-off city administrator, describes himself as conscientious and capable, someone who takes pride in getting the details right.

"I'm a hard worker," he said. "I stay until the job gets done, but a computer will never pick that up."